Criminal Deterrence: A Review of the Literature
Aaron Chalfin
University of Cincinnati
Justin McCrary
UC Berkeley, NBER
May 9, 2014
Abstract
We review economics research regarding the effect of police, punishments, and work on crime, with a particular focus on papers from the
last 20 years. Evidence in favor of deterrence effects is mixed. While
there is considerable evidence that crime is responsive to police and to
the existence of attractive legitimate labor market opportunities, there is
far less evidence that crime responds to the severity of criminal sanctions.
We discuss fruitful directions for future work and implications for public
policy.
JEL Classification: H75, J18, J22, J24, K14, K42
Keywords: crime, deterrence, police, prisons, labor market
Introduction
The day-to-day work of individuals employed in law enforcement, corrections
and other parts of the criminal justice system involves identifying, capturing,
prosecuting, sentencing, and incarcerating offenders. Perhaps the central function of these activities, however, is deterring individuals from participating in
illegal activity in the first place. Deterrence is important not only because it
results in lower crime but also because, relative to incapacitation, it is cheap.
Offenders who are deterred from committing crime in the first place do not
have to be identified, captured, prosecuted, setenced, or incarcerated. For this
reason, assessing the degree to which potential offenders are deterred by either
carrots (better employment opportunities) or sticks (more intensive policing or
harsher sanctions) is a first order policy issue.
The standard economic model of criminal behavior draws on a simple expected utility model introduced in a seminal contribution by the late Gary
Becker. This model envisions crime as a gamble undertaken by a rational individual. According to this framework, the aggregate supply of offenses will
depend on social investments in police and prisons as well as on labor market
opportunities which increase the relative cost of time spent in illegal activities.
Using Becker’s work as a guide, a large empirical literature has developed
to test the degree to which potential offenders are deterred. The papers in this
literature fall into three general categories. First, many papers consider the responsiveness of crime to the probability that an individual is apprehended. This
concept has typically been operationalized as the study of the sensitivity of crime
to police, in particular police manpower or policing intensity. A second group
of papers studies the sensitivity of crime to changes in the severity of criminal
sanctions. The literature assesses the responsiveness of crime to sentence enhancements, three strikes laws, capital punishment regimes and policy-induced
discontinuties in the severity of sanctions faced by particular individuals. The
third group of papers examines the responsiveness of crime to local labor market
conditions, generationally operationalized using either the unemployment rate
or a relevant market wage. This literature seeks to determine whether crime
can be deterred through the use of positive incentives rather than punishments.
The papers in each of these literatures can be viewed as measuring the degree
to which individuals can be deterred from participation in criminal activity.
Each of the literatures is vast. A challenge remains to characterize the pattern of
the empirical findings and explain why individuals appear to be more responsive
(and thus more deterrable) along certain margins than along others. In this
article, we provide a brief review of each of the three literatures introduced above
with the intention of rationalizing several apparently divergent findings. Section
I provides a brief introduction to economic theories of deterrence. Section II
considers research on the effect of police on crime, Section III considers the effect
of prison and/or sanctions on crime and Section IV considers the responsiveness
of crime to local labor market conditions. Section V concludes.
2
1
Theories of Deterrence
Deterrence is an old idea and has been discussed in academic writing at least
as far back as 18th century treatises by Adam Smith (1776), Jeremy Bentham
(1796) and Cesare Beccaria (1798). There are three core concepts embedded in
theories of deterrence — that individuals respond to changes in the certainty,
severity, and celerity (or immediacy) of punishment. Interestingly, in the criminological tradition, deterrence is often characterized as being either general or
specific with general deterrence referring to the idea that individuals respond
to the threat of punishment and specific deterrence referring to the idea that
individuals are responsive to the experience of punishment. Economics prefers
different terminology, reserving the term deterrence for what the criminologist
calls general deterrence and describing specific deterrence as a change in information or, perhaps more exotically, a change in preferences themselves. In
this section, we briefly characterize the way in which economists have formalized these concepts. In general, economic theories of deterrence have focused
more heavily on certainty and severity. However, recent writing has increasingly
characterized deterrence as part of a dynamic framework in which offender behavior is sensitive to their time preferences (Polinsky and Shavell 1999; Lee and
McCrary 2009).
1.1
Economic Models of Crime
The earliest formal model of criminal offending in economics can be found in
Becker’s seminal 1968 paper, Crime and Punishment: An Economic Approach.
The crux of Becker’s model is the idea that a rational offender faces a gamble.
He can either choose to commit a crime and thus receive a criminal benefit
(albeit with an associated risk of apprehension and subsequent punishment) or
not to commit a crime (which yields no criminal benefit but is risk free). The
expected cost of committing a crime is a function of the offender’s probability
of apprehension, p, and the severity of the sanction that he will face upon
apprehension, f . To be more specific, the individual can be said to face three
potential outcomes each of which delivers a different level of utility: 1) the utility
associated with the choice to abstain from crime, Unc , 2) the utility associated
with choosing to commit a crime that does not result in an apprehension, Uc1 ,
and 3) the utility associated with choosing to commit a crime and which results
in apprehension and punishment, Uc2 . In such a formulation, the individual
chooses to commit a crime and only if the following condition holds:
(1 − p)Uc1 + pUc2 > Unc
(1)
That is, crime is worthwhile so long as its expected utility exceeds the utility
from abstention.1
In addition to the clear role played in this model by the probability of apprehension, p, the formulation also suggests the importance of two additional
1 The “if and only if” holds if we maintain that the case of (1 − p)U
c1 + pUc2 =Unc implies
no crime, an unimportant assumption we make henceforth to simplify discussion.
3
exogenous factors that could influence Uc2 and Unc . Crime becomes more attractive when the disutility of apprehension is slight (e.g., less unpleasant prison
conditions), and it becomes less attractive when the utility of work is high (e.g.,
a low unemployment rate or a high wage). Becker operationalizes the disutility
associated with capture using a single exogenous variable, f , which he refers to
as the severity of the criminal sanction upon capture. Typically, f is assumed
to refer to something like a fine, the probability of conviction, or the length of
a prison sentence.2 To a large degree, then, government maintains control over
Uc2 .
The utility associated with abstaining from crime, Unc , is principally a function of the individual’s ability to derive utility from non-illicit activities. In
practice, this is typically thought of as the wage that can be earned in the legal
labor market. When the legal wage rises, Unc rises, thus reducing the relative
benefit of criminal activity. It is fair to say that while government maintains
some control over Unc , it does so to a lesser extent than it does over the utility
of punishment, Uc2 .
Using these ideas, Becker rewrites (fn. 16) the expected utility confronting
an individual contemplating crime as
EU = pU (Y − f ) + (1 − p)U (Y )
(2)
where Y represents the income associated with getting away with crime.3 In
this formulation, crime occurs if and only if EU > Unc . Equivalently, we can
define an indifference point, Y ∗ , such that crime occurs if and only if Y > Y ∗ .
It is easy to see that
p
U (Y ∗ ) − Unc
=
|U (Y ∗ − f ) − Unc |
1−p
(3)
Several important ideas are embedded in (3). First, for the individual to
elect to engage in crime, the gain relative to its loss must exceed the odds
of capture. Dividing the numerator and denominator of the left side by Unc
yields a natural interpretation in terms of percentages. Consider a criminal
opportunity where capture is n times as likely as not. Crime occurs if the
anticipated percent improvement in utility associated with getting away with
it is more than n times as large as the anticipated percent reduction in utility
associated with apprehension. Second, an increase in p unambiguously reduces
the likelihood of crime, as this increases the right-hand side of (3). Third, an
increase in f unambiguously reduces the likelihood of crime as long as U 0 (·) > 0,
as this decreases the left-hand side of (3).
Under risk neutrality, the equation (3) simplifies. Define a as the income
associated with abstaining from crime, i.e., U (a) = Unc , define c = f − b > 0 as
2 In principle, f can be a function of many different characteristics of the sanction including
the length of the sentence, the conditions under which the sentence will be served and the
degree of social stigma that is attached to a term of incarceration, all of which are likely
heterogeneous among the population.
3 As Becker is careful to say, income “monetary and psychic.”
4
the effective cost of punishment, and define Y = a + b and Y ∗ = a + b∗ , where
b is the criminal benefit and b∗ is the criminal benefit at which the individual
is indifferent between crime and abstention. Then equation (3) reduces to
b∗ = c
p
1−p
This simplified version of the Becker model is the starting point of the dynamic
analysis in Lee and McCrary (2009).
A somewhat different focus can be found in Ehrlich (1973), where the notion
of the opportunity cost of engaging in crime is front and center. Perhaps unsurprisingly, labor economists have found it particularly attractive to view crime as
a time allocation choice, and this type of formulation is found in several prominent papers including Lemieux, Fortin, and Frechette (1994), Grogger (1998),
Williams and Sickles (2002) and Burdett, Lagos and Wright (2004) among others.4
The typical time allocation model of crime considers a consumer facing a constant market wage and diminishing marginal returns to participation in crime.
This consumer maximizes a utility function that increases in both leisure (L)
and consumption (C), where consumption is financed by time spent engaged
in either legitimate employment (hm ) at a market wage (w) or time spent in
crime (hc ) with a net hourly payoff of r. The consumer’s constrained optimization problem is to maximize his utility function, U (C, L), subject to the
consumption and time constraints:
C = whm + rhc + I
(4)
L = T − hm − hc
(5)
In (4), consumption is shown to be equal to an offender’s legitimate income plus
his non-legitimate income.5 In (5), T is the individual’s time endowment and
leisure is the remaining time after market work and time spent in crime are
accounted for.6 The parameter r reflects the criminal benefit but it also reflects
the costs of committing crime, namely the risk of capture and the expected
criminal sanction if captured. In other words, r can be thought of as the wage
rate of crime, net of the expected costs associated with the criminal justice
system. In this way, criminal sanctions drive a wedge between the consumer’s
productivity in offending and his market wage, in turn, incentivizing market
work over crime.7
4 For
further details, see Gronau (1980).
represents non-labor income.
6 Grogger assumes that the returns to crime diminish as the amount of time devoted to
criminal activity increases — i.e., there is a function r(·) that translates hours spent participating in crime into income is concave. Diminishing returns implies that those engaging in
criminal activity first commit crimes with the highest expected payoffs (lowest probability of
getting caught and highest stakes) before exploring less lucrative opportunities. However, this
need not be true.
7 In order for an individual to commit any crime at all, there are two necessary and sufficient
conditions. First, the marginal return to the first instant of time supplied to crime must exceed
5I
5
An interesting question in both Becker’s model and Ehrlich’s model is whether
individuals are more deterred by increases in p or in f . Becker addresses this
in a straightforward way by asking whether the expected utility of crime is decreased more by a small percent increase in p or an equivalent percent increase
in f . This makes sense because participation in crime should be monotonic in
its expected utility. Becker’s analysis shows that p is more effective if and only if
U 00 (·) > 0, i.e., if and only if individuals were risk preferring.8 If individuals are
averse to risk, increasing f is more effective than increasing p, and if individuals
are risk neutral, then f and p are equally effective. Becker notes (fn. 12) that
this conclusion is the opposite of that given by Beccaria regarding the effectiveness of punishment versus capture, and that the conclusion is similarly at odds
with contemporaneous views of judges.
The model of Lee and McCrary (2009) emphasizes the dependence of this
conclusion on the time preferences of the individual.9 Intuitively, it seems like
it would be hard to deter an impatient individual using a prison sentence, since
most of the disutility of a prison sentence is borne in the future. Lee and
McCrary propose modeling crime using a modification of the basic job search
model in discrete time with an infinite horizon. Risk-neutrality is assumed,
yet individuals in this model have very different responses to the capture and
punishment.
In their model, criminal opportunities are independent draws from an identical “criminal benefit” distribution with distribution function F (b). The individual learns of an opportunity each period and must decide whether to take
advantage of it. If the individual engages in crime and is caught, she is imprisoned for S periods, where S is an independent draw from an identical sentence
length distribution. As in the Becker model, capture occurs with probability
p. If the individual abstains from crime, she obtains flow utility a and faces
the same problem the next period. If she commits crime and is not caught, she
obtains flow utility a + b and faces the same problem next period. Finally, if
she commits crime and is caught, she obtains flow utility a − c for S periods,
before confronting the same problem at the conclusion of her sentence.
The criminal benefit at which the individual is indifferent between crime and
abstention is given by
Z ∞
p
p
+ν c
+p
(1 − F (z))dz
(6)
b∗ = c
1−p
1−p
b∗
PS−1
where ν = E[ s=1 δ s ] is a summary parameter governing how the distribution
the individuals valuation of time (in terms of how much consumption the person would be
willing to forego for more time) when all time is devoted to non-market, non-crime activities.
Second, the marginal return to crime for the first crime committed must exceed the individuals
market wage. Thus, those who can command high wages or those who place very high value
on time devoted to non-market/non-criminal uses will be the least likely to engage in criminal
activity.
8 Note, however, the observant criticism of Brown and Reynolds (1973), showing that this
clean conclusion is the result of the modeling assumption that the baseline utility is that of
getting away with crime.
9 Further details regarding the Lee and McCrary model are given in McCrary (2010).
6
of sentences affects decision-making and δ is the discount factor.10,11
As in Becker’s static model, crime is reduced by increases in p and increases
in c. The added feature of this model, however, is that crime is also reduced
by increases in sentence lengths, and this behavioral mechanism is modulated
by time preferences. As is intuitive, patient individuals are quite responsive to
increases to sentence lengths, but impatient individuals show much more muted
responses. In the limit as the discount factor approaches zero, the individual
is arbitrarily more responsive to capture than to punishment. The Lee and
McCrary model thus provides a simple way to re-introduce older ideas regarding
the importance of celerity into a Becker model.12
Ultimately, the models proposed by Becker and Ehrlich yield three main
behavioral predictions: 1) the supply of offenses will fall as the probability of
apprehension rises, 2) the supply of offenses will fall as the severity of the criminal sanction increases and 3) the supply of offenses will fall as the opportunity
cost of crime rises. In other words, behavioral changes can be brought about either using carrots (better employment opportunities) or sticks (criminal justice
inputs). The following section connects these core predictions to the empirical
literatures that have sought to test whether these predictions hold in the real
world.
1.2
Deterrence vs. Incapacitation
Generally speaking there are two mechanisms through which criminal justice
policy reduces crime: deterrence and incapacitation. When by virtue of a policy
change individuals elect not to engage in crime they otherwise would have in
the absence of the change, we speak of the policy deterring crime. On the other
hand, a policy change may also take offenders out of circulation, as for example
with pre-trial detention or incarceration, preventing crime by incapacitating
individuals. The incapacitation effect can be thought of as the mechanical
response of crime to changes in criminal justice inputs. While deterrence can
arise in response to any policy that changes the costs or benefits of offending,
incapacitation arises only when the probability of capture or the expected length
of detention increases.
The existence of incapacitation effects has profound implications for the
study of deterrence. In particular, while research that considers the effect of a
change in the probability of capture will, generally speaking, identify a mixture
of deterrence and incapacitation effects, research that considers changes in the
opportunity cost of crime is more likely to isolate deterrence. Likewise, while
10 For example, if we take S to be geometric, i.e., S has support 1, 2, . . . , ∞ and P (S = s) =
q(1 − q)s−1 , where q is the per-period release probability, then standard results using infinite
series show that ν = (1 − q)δ/(1 − (1 − q)δ). Interestingly, this shows that under a geometric
distribution for sentence lengths, reducing the probability of release is equivalent to increasing
the individual’s patience.
11 Equation (6) is not an explicit equation for b∗ , but it can be viewed as defining an implicit
function. We can use numerical methods to solve for b∗ (e.g., Newton’s method works well),
and comparative statics are straightforward using the implicit function theorem.
12 For related modeling ideas from criminology, see Nagin and Pogarsky (2000), for example.
7
research on the effect of sanctions typically results in a treatment effect that is
a function of both deterrence and incapacitation, clever research designs have
been used to identify the effect of an increase in the severity of a sanction that
is unlikely to result in an immediate increase in incapacitation.
For each literature discussed in this paper, we provide a discussion of the
degree to which empirical estimates can be interpreted as providing evidence
of deterrence as distinct from incapacitation. However, it is important to note
that deterrence is itself a black box. In order to empirically observe a behavioral
response of crime to a particular policy level, it must be the case that potential
offenders perceive that the cost of committing a crime has changed (Nagin 1998;
Durlauf and Nagin 2011; Nagin 2013). Moreover, the behavioral response of
crime will depend on the accuracy of those perceptions. To wit, an intervention
that successfully convinces potential offenders that the expected cost of crime
has increased, regardless of whether this is actually the case, will likely reduce
crime. The challenge for cost-effective public policy is to optimally trade off
between police and prisons so as to maximize perceptual and, as such, actual
deterrence.
2
Police and Crime
Becker’s prediction that the aggregate supply of crime will be sensitive to society’s investment in police arises from the idea that an increase in police presence,
whether it is operationalized through increased manpower or increased productivity, increases the probability that an individual is apprehended for having
committed a particular offense. To the extent that potential offenders are able
to observe an increase in police resources and perceive a correspondingly higher
risk to criminal participation, crime should decline through the deterrence channel.
Empirically, the challenge for this literature is that changes in the intensity of
policing are generally not random. As a result, it is difficult to identify a causal
effect of police on crime using natural variation in policing. Conceptually, the
issue is that the responsiveness of crime to police may also reflect an important
role for incapacitation. This arises from the idea that police tend to reduce
crime mechanically, even in the absence of a behavioral response, by arresting
offenders who are subsequently incarcerated and incapacitated.13 The extent
to which investments in police are cost-effective depends, in large part, on the
degree to which police deter rather than simply incapacitate offenders. In this
section, we consider the responsiveness of crime to both police manpower and
police tactics, broadly defined. For each literature we discuss the challenges
with respect to both econometric identification as well as interpretation of the
resulting parameters as evidence in favor of deterrence.
13 In this context, deterrence can arise either from a general decrease in offending or from a
shift towards less productive but correspondingly less risky modes of offending — for example,
a shift from robbery to larceny.
8
2.1
Police Manpower
A large literature has used city- or state-level panel data and, recently, a variety
of quasi-experimental designs to estimate the elasticity of crime with respect
to police manpower.14 This literature is ably summarized by Cameron (1988),
Nagin(1998), Eck and Maguire (2000), Skogan and Frydal (2004) and Levitt
and Miles (2006), all of whom provide extensive references.
The early panel data literature tended to report small elasticity estimates
that were rarely distinguishable from zero and sometimes even positive, suggesting perversely that police increase crime.15 The ensuing discussion in the
literature was whether police reduce crime at all. Beginning with Levitt (1997),
an emerging quasi-experimental literature has argued that simultaneity bias is
the culprit for the small elasticities in the panel data literature.16 The specific
concern articulated is that if police are hired in anticipation of an upswing in
crime, then there will be a positive bias associated with regression-based strategies, masking a true negative elasticity. The recent literature has therefore
generally focused instead on instrumental variables (IV) strategies designed to
overcome this bias.
The first plausible instrumental variable to study the effect of police manpower on crime was proposed by Levitt (1997). Leveraging data on the timing
of mayoral and gubernatorial elections, Levitt provides evidence that in the year
prior to a municipal or state election, police manpower tends to increase, presumably due to the desire of elected officials to appear to be “tough on crime.”
The exclusion restriction is that but for increases in police manpower, crime
does not vary cyclically with respect to the election cycle. Using data from
57 cities spanning 1972-1997, Levitt reports very small least squares estimates
of the effect of police and crime that are consistent with the prior literature.
However, IV estimates are large and economically important with elasticities
ranging from moderate in magnitude for property crimes (-0.55 for burglary
and -0.44 for motor vehicle theft) to large in magnitude for violent crimes such
as robbery (-1.3) and murder (-3). Ultimately, following a reanalysis of the data
by McCrary (2002), the IV coefficients reported by Levitt were found to be
insignificant after a problem with weighting was addressed.
Levitt (1997) has given rise to a series of related papers that seek to identify a national effect of police manpower on crime by identifying conditionally
exogenous within-city variation in police staffing levels. These papers include
Levitt (2002) which uses variation in firefighter numbers as an instrument for
police manpower, Evans and Owens (2007) who instrument for police manpower
14 This elasticity can be thought of as a reduced form parameter that captures both deterrence effects as suggested by neoclassical economic theory as well as incapacitation effects
that arise when offenders are incarcerated and thus constrained in their ability to offend.
15 Papers in this literature employ a wide variety of econometric approaches. Early empirical
papers such as Ehrlich (1973) and Wilson and Boland (1977) focused on the cross-sectional
association between police and crime.
16 Some of the leading examples of quasi-experimental papers are Levitt (2002), DiTella and
Schargrodsky (2004), Klick and Tabarrok (2005), Evans and Owens (2007), Lin (2009) and
Machin and Marie (2011).
9
using the size of federal COPS grants awarded to cities to promote police hiring
and Lin (2009) who instruments for changes in police manpower using the idea
that U.S. states have differential exposure to exchange rate shocks depending on
the export intensity of local industry. These strategies consistently demonstrate
that police do reduce crime.17 However, the estimated elasticities display a wide
range, roughly -0.1 to -2, depending on the study and the type of crime. Moreover, relatively few of the estimated elasticities are significant at conventional
levels of confidence reflecting a great deal of sampling variability and the use
of relatively weak instruments. In many cases, extremely large elasticities (i.e.,
those larger than 1 in magnitude) cannot be differentiated from zero. Overall,
Chalfin and McCrary (2013) characterize the pattern of the cross-crime elasticities as, in general, favoring a larger effect of police on violent crimes than on
property crimes with especially large effects of police on murder, robbery and
motor vehicle theft.18
A second noteworthy contribution to the modern police manpower literature is that of Marvell and Moody (1996) who leverage the concept of Granger
causality to explore the extent to which police manpower is, in fact, responsive
to changes in crime. The motivation behind such an approach is that if crime is
responsive to lagged police but police staffing is not responsive to lagged crime,
then the case for instrumental variables is weakened considerably. Finding no
evidence of a link between lagged crime rates and current police staffing levels
at either the state or city level, Marvell and Moody estimate the responsiveness of crime to police using a standard two-way fixed effects model and report
elasticities that are fairly small in magnitude (ranging from -0.15 for burglary
to -0.30 for motor vehicle theft) and are more consistent with the early least
squares literature than the IV literature that has proliferated in recent years.
Ultimately the Granger causality exercise is subject to the same omitted variables bias issues that plague any least squares regression model and is therefore
of dubious value in establishing causality. Nevertheless, the weak evidence of a
link between lagged crime and current police staffing presented in Marvell and
Moody is, in our view, underappreciated. Given the large discrepancy between
Marvell and Moody’s estimates and those in Levitt (1997) which use the same
underlying data, one of two propositions must be true: 1) Marvell and Moody’s
estimates of the effect of lagged crime on police manpower are biased due to
the exclusion of important omitted variables, 2) There is no simultaneity bias
between police and crime; discrepancies between least squares and IV estimates
are instead driven by measurement errors in either police staffing or measures
of UCR index crimes. This is an idea that is dealt with in detail in Chalfin and
McCrary (2013). Leveraging two potentially independent measures of police
manpower (one from the FBI’s Uniform Crime Reports and another from the
17 Notably, Worrall and Kovandzic (2007) report no reduced form relationship between
COPS grants and crime. However, their analysis is based on a smaller sample of cities than
the analysis of Evans and Owens (2007).
18 This pattern is found in several prominent panel data papers, in particular Levitt (1997),
Evans and Owens (2007) and Chalfin and McCrary (2013) each of which report especially
large elasticity estimates for murder (-0.6 to -0.8) and robbery (-0.5 to -1.4).
10
U.S. Census’ Annual Survey of Government Employment) for a sample of 242
U.S. cities over a 51 year time period, Chalfin and McCrary construct measurement error corrected IV models using one measure of police as an instrument
for the other. Their principal finding is that elasticities reported in the recent
IV literature can be replicated by simply correcting for measurement errors in
police data and without explicitly addressing the possibility of simultaneity bias.
The resulting implication is that Marvell and Moody’s basic inference regarding
the lack of causality running from crime to police manpower may be correct.
A related contribution in Chalfin and McCrary is to estimate police elastcities with remarkable precision, reporting elasticities of −0.67 ± 0.48 for murder,
−0.56 ± 0.24 for robbery, −0.34 ± 0.20 for motor vehicle theft and −0.23 ± 0.18
for burglary.
While the majority of the police manpower literature uses aggregate data,
there is a corresponding literature that assesses the impact of police on crime
using natural experiments in a particular jurisdiction. An early account of such a
natural experiment is found in Andenaes (1974) who documents a large increase
in crime in Nazi-occupied Denmark after German soldiers dissolved the entire
Danish police force (Durlauf and Nagin 2011; Nagin 2013). Modern literature
has found similarly large effects. In particular, DeAngelo and Hansen (2010)
document an increase in traffic fatalities that occured in the aftermath of a
budget cut in Oregon that resulted in a mass layoff of state troopers. Similarly
Shi (2009) reports an increase in crime in Cincinnati, OH in the aftermath of an
incident in which police used deadly force against an unarmed African-American
teenager.19
2.2
Police Deployment and Tactics
The police manpower literature is informative with respect to the aggregate response of crime to increases in police staffing. However, the aggregate manpower
literature leaves many interesting and important questions unanswered. In particular, to what extent do the estimated elasticities reflect deterrence? Likewise,
what is the specific mechanism that leads to deterrence — if the mechanism is
based on perceptual deterrence — the idea that offenders observe an increase in
police presence and adjust their behavior accordingly — then it should be the
case that offending is especially sensitive to large and easily observed changes
in police deployment and tactics. To address these questions, a related literature that is found mostly in criminology has studied the effect of changes in
the intensity of policing on crime with a distinct focus on the crime-reducing
effect of various “best practices.” In particular, declines in crime that are not
attributable to spatial displacement have been linked to the adoption of “hot
spots” policing (Sherman and Rogan 1995, Sherman and Weisburd 1995, Braga
2001, Braga 2005, Weisburd 2005, Braga and Bond 2008, Berk and MacDonald
2010), “problem-oriented” policing (Braga et al 1999, Braga, Kennedy, Waring and Piehl 2001, Weisburd, Telep, Hinckle and Eck 2010) and a variety of
19 As Shi (2009) notes, the police response to the riot was to reduce productivity disproportionately in riot-affected neighborhoods.
11
other proactive approaches. Similarly, a large research literature that has examined the local impact of police crackdowns has consistently found large and
immediate (but typically not lasting) reductions in crime in the aftermath of
hyper-intensive policing (Sherman 1990). Such findings are further supported
by evidence from several informative natural experiments which have identified plausibly exogenous variation in the intensity of policing. Three prominent
examples are Klick and Tabarrok (2005) who study the effect of police redployments in Washington DC that result from shifts in terror alert levels, DiTella
and Schargrodsky (2004) who study the effect of a shift in the intensity of policing in certain areas of Buenos Aires after a 1994 synagogue bombing and Draca,
Machin and Witt (2011) who study police redployments in the aftermath of the
2005 London tube bombings.
The literature on police deployments and tactics has focused predominantly
on three types of interventions. The first is an innovation commonly referred to
as “hot spots” policing. As the moniker suggests, hot spots policing describes
a strategy in which police are disproportionately deployed to areas in a city
that appear to attract disproportionate levels of crime.20 The second type of
intervention is broadly referred to as “problem oriented” policing. This term is
used broadly and refers to a collection of focused deterrence strategies that are
designed to change the behavior of specific types of offenders. A final intervention that has received attention in the literature is that of “proactive” policing.
Proactive policing refers to strategies that are deigned to make policing more
intensive, holding resources fixed. The idea can be traced back to the concept of
“broken windows” policing introduced by Wilson and Kelling (1982) and refers
to the notion that just like fixing a broken window sends a message to would-be
vandals that the community cares about maintaining social order, arresting individuals for relatively minor infractions sends a message to potential offenders
that the police are watchful.
2.2.1
Hot Spots Policing
We begin with a discussion of hot spots policing. As obvious as the idea may
sound, given resource constraints, the feasibility of such a deployment strategy
relies on crime being sufficiently concentrated in a relatively small number of hot
spots. A seminal paper by Sherman, Gartin and Buerger (1989) is the first to
provide descriptive data on the degree to which crime is spatially concentrated.
Using data from Minneapolis, Sherman and co-authors found that just three
percent of addresses and intersections in Minneapolis produced 50 percent of
all calls for service to to the police. This finding is echoed by Weisburd, Maher
and Sherman (1992) and in more recent papers by Weisburd et al (2004) and
Weisburd, Morris and Groff (2009) which report that a very small percentage
of street segments in Seattle accounted for 50 percent of crime incidents for
20 The idea that crime hot spots might exist is an old one and can be found at least as far
back as Shaw and McKay (1942). Modern research has linked criminal activity to specific
types of places such as bars (Roman and Reid 2010) and apartment buildings as well as to
places that lack formal or informal guardians (Eck and Weisburd (1995).
12
each year over a fourteen year period.21 Naturally, the observation that crime
is so highly concentrated in a very small number of places has led to efforts to
intensify the focus of police resources on these places. These interventions have,
in turn, led to a corresponding experimental and quasi-experimental research
literature that seeks to evaluate the efficacy of such strategies.
The first order question that the hot spots policing literature seeks to address
involves the degree to which highly localized crime is responsive to a change in
the intensity of policing. However, a particular feature of this research makes
it particularly salient for the study of deterrence (Nagin 2013). Notably, while
the literature tends to find that intensive policing reduces crime, elements of
intensive policing such as rapid response times do not appear to increase the
likelihood of an arrest (Spelman and Brown 1981). Such a pattern in the data
tends to be consistent with deterrence but not with incapacitation.
The first test of policing crime hot spots may be found in a 1995 randomized
experiment conducted by Sherman and Weisburd in Minneapolis. The experiment tested whether doubling the intensity of police patrols in crime hotspots
resulted in a decrease in crime and found that crime declined by approximately
10 percent in experimental places relative to control places. No evidence of
crime displacement — the idea that crime is merely moved to an adjacent place
— was found. Findings in Sherman and Weisburd (1995) have, to a large extent,
been replicated in other places and contexts including the presence of open air
drug markets and “crack’ houses” (Hope 1994; Weisburd and Green 1995; Sherman and Rogan 1995b), violent crime hot spots (Sherman and Rogan 1995a;
Braga et al 1999; Caeti 1999) and places associated with substantial social disorder (Braga and Bond 2008; Berk and MacDonald 2010). Indeed a review of
the literature by Braga (2008) identified nine experiments or quasi-experiments
involving hot spots policing and noted that seven of the nine studies found
evidence of significant crime reductions. Notably, a majority of the literature
finds no evidence of displacement of crime to adjacent neighborhoods indicating
that the data are more consistent with deterrence (Weisburd et al 2006) while
a number of studies have found that the opposite is true — that there tends to
be a diffusion of benefits to non-treated adjacent places (Sherman and Rogan
1995a; Braga et al 1999; Caeti 1999).22
2.2.2
Problem-Oriented Policing
Intensive policing of hot spots is one way that police potentially deter crime.
Another broad deterrence-based strategy is that of problem-oriented policing.
Broadly speaking, this strategy entails engaging with community residents to
identify the most salient local crime problems and designing strategies to deter
unwanted behavior. Undoubtedly the most well-known evaluation of a problemoriented policing approach is that of Boston’s Operation Ceasefire by Kennedy,
21 An excellent review of this literature may be found in Weisburd, Bruinsma and Bernasco
(2009).
22 An excellent review of the theory and empirical findings regarding displacement in this
literature can be found in Weisburd et al (2006).
13
Braga, Piehl and Waring (2001). The stated purpose of Ceasefire was to reduce
youth gun violence in Boston, MA. The intervention involved a multi-faceted
approach and included efforts to disrupt the supply of illegal weapons to Massachusetts as well as messages communicated by police directly to gang members
that authorities would use every available “lever” to punish gangs collectively
for violent acts committed by individual gang members. In particular, police
indicated that the stringency of drug enforcement would hinge on the degree
to which gangs used violence to settle business disputes. The result of the intervention was that youth violence fell considerably in Boston relative to other
U.S. cities.
Indeed Ceasefire was so successful that it has given rise to a number of
similarly-motivated strategies that are collectively referred to as “pulling levers.”
Prominent evaluations of pulling levers interventions include research carried
out in Richmond, VA (Raphael and Ludwig 2003), Indianapolis (McGarrell at
el 2006), Chicago (Papachristos, Meares and Fagan 2007), Stockton, CA (Braga
2008b), Lowell, MA (Braga et al 2008), High Point, NC (Corsaro, Hunt, Hipple
and McGarrell 2012), Newark, NJ (Boyle et al 2010), Nashville (Corsaro and
McGarrell 2010), Cincinnati (Engel, Corsaro and Tillyer 2010) and Rockford,
IL (Corsaro, Brunson and McGarrell 2010). Researchers have also evaluated a
multi-city pulling levers strategy known as Project Safe Neighborhoods (PSN)
which enlisted to cooperation of federal prosectuors to crack down on gun violence.
On the whole, evaluations of pulling levers strategies produce extremely
promising results though inference is invariably complicated by a lack of randomized experiments and the inherent difficulty in identifying appropriate comparison cities. With respect to individual evaluations, reductions in crime have
been found in High Point, Chicago, Indianapolis, Stockton, Lowell, Nashville
and Rockford and null findings have been found in Richmond and Cincinnati.23
With respect to Project Safe Neighborhoods, the research is promising but not
definitive. While McGarrell, Corsaro, Hipple and Bynum (2010) report that
declines in crime were greater in PSN cities than in non-PSN cities, Corsaro,
Chalfin and McGarrell (2013) point out that when pre-intervention trends are
more fully accounted for, average declines in crime continue to exist albeit with
a great deal of heterogeneity among cities.
2.2.3
Proactive Policing
A final strand of the police tactics literature in criminology investigates the responsiveness of crime to the intensity of policing, holding resources constant, an
idea is that is generally referred to as “proactive” policing. As there is no standardized way to assess the extent to which individual police departments engage
in policework that is proactive, in practice, this literature seeks to understand
if the intensity of arrests for minor infractions has an effect on the incidence of
more serious crimes. Such an empirical operationalization was first proposed
23 Braga and Weisburd (2012) provided an excellent review of the literature including a
comprehensive meta-analysis of the research findings.
14
by Sampson and Cohen (1988) and has been replicated to various degrees by
MacDonald (2002) and Kubrin, Messner, Deane, McGeever and Stucky (2005).
The general strategy is to regress crime rates on a measure of policing intensity. In practice, policing intensity has been operationalized using the number of
DUI and disorderly conduct arrests made per police officer. Using this approach
has, in some cases, led to findings that are consistent with a deterrence effect
of proactive policing. However, in the best controlled models, coefficients on
the proactive policing proxy become small and insignificant. More importantly,
these models are plagued by problems of simultaneity bias, omitted variables
and the inevitable difficulty involved in finding a credible proxy for the concept of proactive policing as opposed to simply an environment that is rich in
opportunities for police officers to make arrests.
A second focus of the literature has been on the advent of “broken windows”
policing (also known as “order maintenance policing,” a policy innovation proposed by Wilson and Kelling (1982). The idea behind broken windows policing
is that police can affect crime through tough enforcement of laws governing
relatively minor infractions such as vandalism and turnstyle jumping. Broken
windows policing, in theory, operates primarily through perceptual deterrence
— if offenders observe that police are especially watchful, they may update their
perceived probability of apprehension for a more serious crime and accordingly
will decrease their participation in crime. In the popular media, broken windows policing is an idea that is heavily associated with Mayor Rudolph Giuliani
and New York Police Commissioner William J. Bratton who has attributed the
dramatic decline in crime in New York City after 1990 to its rollout. A corresponding research literature has arisen to evaluate the effectiveness of broken
windows policing — in practice, this literature has focused disproportionately
on the experience of New York City. This literature produces mixed findings.
On the one hand, time series analyses by Kelling and Sousa (2001) and Corman
and Mocan (2005) find that misdemeanor arrests are negatively associated with
future arrests for more serious crimes such as robbery and motor vehicle theft.
However, later research has pointed out that these studies omit a control group
and has tended to find that New York’s aggregate crime trends are similar to
those of other cities that did not institute a policy of broken windows policing
(Eck and Maguire 2000; Rosenfeld, Fornango and Baumer 2005; Harcourt and
Ludwig 2006). With respect to the newest research, a 2007 paper by Rosenfeld,
Fornango and Rengifo uses precinct-level data and finds evidence of a negative
effect of broken windows policing on crime, albeit a very small effect. Similarly,
a particularly careful paper by Caetano and Maheshri (2013) finds no evidence
of an effect of “zero tolerance” law enforcement policies on crime using microdata from Dallas.24 Overall, our reading if this literature is that the evidence
in favor of an important effect of proactive policing on crime is weak.
24 Broken
windows policing and the associated “stop and frisk” policy implemented by the
New York City police department has generated substantial public controversy. A 2009 paper
by Fagan Gellar, Davies and West provides evidence of the demographic burden of such policies
which is disproportionately borne by African-Americans.
15
2.2.4
Police Deployments
The ubiquity of the hot spots, problem-oriented and proactive policing literatures in criminology has spawned a parallel literature in economics that seeks to
learn from natural experiments in police deployments. Three prominent studies
are those of DiTella and Schargrodsky (2004), Klick and Tabarrok (2005) and
Draca, Machin and Witt (2011). Each of these studies leverages a redeployment
of police in response to a perceived terrorist threat. The appeal of these studies is that terrorist threats are plausibly exogenous with respect to trends in
city-level crime and therefore represent a unique opportunity to learn about the
response of crime to changes in normal routines of policing. DiTella and Schargrodsky study the response of police in Buenos Aires to the 1994 bombing of a
Jewish community center. In the aftermath of the bombing, Argentine police
engaged in a strategy of “target hardening” synagogues by deploying disproportionate numbers of officers to blocks with synagogues or other buildings housing
Jewish organizations. DiTella and Schargordsky report that the intervention
led to a large decline in motor vehicle thefts on the blocks that received additional police patrols though the effects. Notably this result has been called into
question by Donohue, Ho and Leahy (2013) who reanalyzed the original data
and report evidence that is more consistent with spatial displacement of crime
rather than crime reduction. However, with respect to identifying behavioral
changes among offenders, both stories are equally consistent with deterrence.
In a similar study, Klick and Tabarrok (2005) utilize the fact that when terror
alert levels set by the U.S. Department of Homeland Security rise, property
crime (but not violent crime) tends to fall in Washington D.C., with especially
large declines in areas that receive the largest redployments of police protection. With respect to the U.K., Draca, Machin and Witt (2011) study the 2005
London tube bombings which resulted in sizeable shifts in the deployments of
police from the surburbs to central London and find that “street crimes” such
as robbery and theft are reduced considerably in areas that received additional
officers.
With respect to studying variation in the spatial concentration of police, two
additional papers are worth noting. Cohen and Ludwig (2002) exploit shortterm variation in the intensity of police patrols by day of the week in several
different Pittsburgh patrol areas. They found that shootings were considerably lower in areas and on days that received more intensive police patrols.
With respect to the long-term consequences of patterns of police deployments,
MacDonald, Klick and Grunwald (2013) use a spatial regression discontinuity
design to study the impact of especially intensive policing around the University
of Pennsylvania, a large urban university campus. In particular, areas directly
adjacent to the university received police patrols from both the university and
municipal police. Areas slightly further away from the campus received only municipal police patrols. The finding is that street crimes are substantially higher
in the blocks just outside the area patrolled by the university police relative to
the blocks just inside the university patrol area.
16
2.3
Deterrence vs. Incapacitation
The literature has reached a consensus that increases in police manpower reduce
crime, at least for a population-weighted average of U.S. cities. With respect
to police deployments and tactics, the literature supports the idea that crime is
responsive to hot spots policing and pulling levers strategies while evidence in
favor of an effect of proactive policing strategies such as broken windows and
disorder policing is more suspect. A remaining issue is to address the degree
to which each of these literatures is informative with respect to disentangling
deterrence from incapacitation.
With respect to the aggregate manpower literature, Levitt (1998) provides
the first attempt to systematically unpack the relationship between deterrence
and incapacitation by empirically examining the link between arrest rates and
crime, a relationship which is negative. Levitt posits that this negative relationship can be explained either by deterrence, incapacitation or measurement errors
in crime. Ruling out measurement errors as a likely culprit, he differentiates between deterrence and incapacitation using the effect of changes in the arrest
rate for one crime on the rate of other crimes.25 As Levitt notes, “in contrast
to the effect of increased arrests for one crime on the commission of that crime,
where deterrence and incapacitation are indistinguishable, it is demonstrated
that these two forces act in opposite directions when looking across crimes. Incapacitation suggests that an increase in the arrest rate for one crime will reduce
all crime rates; deterrence predicts that an increase in the arrest rate for one
crime will lead to a rise in other crimes as criminals substitute away from the
first crime.” Levitt concludes that deterrence appears to be the more important
factor, particularly for property crimes. Owens (2013) reports a similar finding,
examining whether variation in police staffing resulting from the COPS hiring
program led to increased arrests. Despite the fact that the program appears to
have led to a decline in crime, no effect is found on arrests. As a result, Owens
concludes that there is little evidence in favor of incapacitation, thus implying
a large role for deterrence.
While analyses by Levitt (1998) and Owens (2013) are suggestive of a meaningful role for deterrence, it is nevertheless difficult to disentangle deterrence
from incapacitation in this way. In particular, a null relationship between police and arrests is also consistent with the idea that police productivity decreases
when there are fewer crimes to investigate. For this reason, while the aggregate
data literature is ideal for understanding the overall relationship between police
and crime, it is only somewhat informative with respect to the magnitude of
deterrence. This point is further compounded by the observation that research
has yet to document the degree to which offenders perceive or are aware of
increases in police manpower (Nagin 1998).
We suspect that the literature on police tactics is considerably more infor25 Utilizing an insight from Grilliches and Hausman (1986) — that measurement errors
should yield the greatest bias in short-differenced regressions — Levitt (1998) compares regression estimates of the relationship between crime and arrest rates using short- and longdifferences, finding similar effects.
17
mative with respect to identifying deterrence. In particular, offenders are more
likely to be aware of an enhanced police presence in small, local areas than
relatively small changes in the number of police in a city spread out over a
large geographic area. Likewise, while offenders tend to commit crimes locally,
in order for incapacitation to explain the large declines in crime that occur in
hot spots, it would have to be the case that offending is so local so as to be
specific to a group of one or two blocks. The large drops in crime that occur
in crime hotspots after they are more aggressively policed is more consistent
with deterrence than with incapacitation. Focused deterrence (“pulling levers”)
strategies are also particularly informative in that declines in crime have been
shown to be specific to the focus of the intervention. To the extent that at
least some offenders are generalists rather than specialists who commit only a
certain type of offense, such a pattern is more consistent with deterrence than
with incapacitation.
In sum, while it remains possible that an increased police presence lowers
crime by situating police officers in locations where they are more likely to arrest
and incapacitate potential offenders, on the whole, the high degree of visibility
around police crackdowns or hot spots policing suggests a potentially greater
role for deterrence.26
3
Sanctions and Crime
A second idea in Becker’s neoclassical model of offending is that crime will be
responsive to the severity of punishment. Accordingly, a parallel literature considers the responsiveness of crime to the severity of criminal sanctions, along
both the intensive and extensive margin. Three literatures, in particular, are
worth mentioning. First, a series of papers considers the effect of sentencing
policy generally or, alternatively, sentence enhancements on crime to test the
prediction that crime will decrease in response to a sanction regime that either raises the probability of a prison sentence or raises the length of a prison
sentence if given. A corresponding literature considers the effect of laws that
govern the age of criminal majority and, as such, generate large and pervasive
discontinuities in the sanctions that individual offenders face. Finally, a particularly prominent literature considers the effect of a capital punishment regime or
the incidence of executions on murder. Since executions enhance the expected
severity of the sanction without directly affecting an offender’s probability of
capture, this literature is potentially compelling with respect to understanding
deterrence as, subject to satisfying the standards of econometric identification,
it allows for the isolation of a pure deterrence effect.
26 An important exception to this intuition however can be found in Mastrobuoni (2013) who
studies the responsiveness of crime to regular shift changes among the various police forces in
Milan, Italy. Matsrobuoni finds that despite large temporal discontinuities in clearance rates
during shift changes, robbers do not appear to exploit these opportunities and oncludes that
there is only limited evidence of deterrence. A remaining question is the extent to which the
result depends on the ability of potential offenders to accurately perceive these discontinuities.
18
3.1
Sentencing
One of the most basic tests of the neoeconomic model of crime concerns the
responsiveness of crime to the severity of criminal sanctions. Over the past
few decades, a literature has arisen to document the sensitivity of crime to
various sentencing schemes, sentence enhancements, clemency policies,“three
strikes” laws and other legislative actions which change the expected cost of
a criminal sanction. A corresponding literature measures the responsiveness
of crime to the size of the prison population. With respect to identification,
two challenges are particularly pressing. First, it is difficult to discern the
effect of sentencing policies (which, in the United States, are generally enacted
at the state level) from other crime reduction interventions as well as timevarying factors that inform the supply of crime more generally. Second, just
as prison populations may affect crime, crime may have a reciprocal effect on
prison populations creating the potential for simultaneity bias.27 With respect
to identfying deterrence, the chief difficulty is that sanctions lead to deterrence
but typically also to incapacitation. This section reviews the literature that
seeks to understand the relationship between sanctions and offending with a
particular interest in discerning the effect that sanctions have on deterrence.
3.1.1
Prison Populations and Crime
While identifying the elasticity of crime with respect to a sanction, in principle,
requires an exogenous shock to the sanctions regime, a natural starting point
in unraveling the crime-sanctions relationship is to consider the elasticity of
crime with respect to the size of the prison population. Studies of the crimeprison population elasticity generally utilize state-level panel data and regress
the growth rate in crime on the first lag of the growth rate in a state’s share of
prisoners. Marvell and Moody (1994) provide the first credible empirical investigation of the elasticity of crime with respect to prison populations, estimating
an elasticity of -0.16. As in their police paper, they use the concept of Granger
causality in an attempt to rule out a causal relationship that runs from crime
to prison populations. As discussed in Section 2, the approach does not offer
particularly strong reasons to believe in the ignorability of selection bias.
A genuine breakthrough in this literature is found in Levitt (1996) who, using
similar data, exploits exogenous variation in state incarceration rates induced
by court orders to reduce prison populations. The intuition behind the approach
is that the timing of discrete reductions in a state’s prison population owing to
a court order should be as good as random. Levitt’s estimated elasticities are
considerably larger than those in Marvell and Moody: -0.4 for violent crimes
and -0.3 for property crimes, while the largest elasticity reported is for robbery
(-0.7).28 An alternative identification strategy can be found in Johnson and
Raphael (2012) who develop an instrumental variable to predict future changes
27 For a comprehensive review of identification issues in this literature see Durlauf and Nagin
(2011).
28 Levitt’s analysis is replicated by Spelman (2000) who reports qualitatively similar findings.
19
in incarceration rates. The instrument is constructed by computing a theoretically predicted dynamic adjustment path of the aggregate incarceration rate in
response to a given shock to prison entrance and exit transition probabilities.
Given that incarceration rates adjust to permanent changes in behavior with
a dynamic lag, the authors identify variation in incarceration that is not due
to contemporaneous criminal offending. Using state level panel data covering
1978-2004, Johnson and Raphael (2012) estimate the elasticity of crime with respect to prison populations of approximately -0.1 for violent crimes and -0.2 for
property crimes. Notably the estimated elasticities for earlier time periods were
considerably larger and closer in magnitude to those estimated by Levitt (1996).
Johnson and Raphael conclude that the criminal productivity of the marginal
offender has changed considerably over time as incarceration rates have risen, a
conclusion that is echoed by Liedka, Piehl and Useem (2006). With respect to
juveniles, Levitt (1998) studies the response of juvenile crime to the punitiveness of state-level juvenile sentencing along the extensive margin (the number of
juveniles in custody per capita), concluding that changes in juvenile sentencing
explain approximately 60 percent of the growth in juvenile crime during the
1970s and 1980s. Using Levitt’s results, Lee and McCrary (2009) compute an
implied elasticity for violent crimes of -0.4.
In sum, estimates of the elasticity of crime with respect to prison are generally modest and fall between -0.1 and -0.7. Estimates for violent and property
crimes are of approximately equal magnitude and there is evidence that the elasticity has diminished considerably over time as prison populations have grown.
Our best guess is that the current elasticity of crime with respect to prison
populations is approximately -0.2 as reported by Johnson and Raphael.29 This
funding is further bolstered by a recent evaluation of “realignment,” a policy
implemented in California to reduce prison overcrowding by sending additional
inmates to county jails where they tend to serve shorter sentences. Lofstrom
and Raphael (2013) report that, with the exception of motor vehicle theft, there
is no evidence of an increase in crime despite the fact that 18,000 offenders who
would have been incarcerated are on the street due to the realignment policy.
The magnitude of this elasticity leaves open the possibility for non-trivial deterrence effects of prison but, given that prison generates sizeable incapacitation
effects, the magnitude of deterrence is likely small.
3.1.2
Shocks to the Sanctions Regime
A related literature considers the effect of a discrete change in a jurisdiction’s
sanctions regime that is plausibly not a function of crime trends more generally. The general approach is to utilize a differences-in-differences design to
compare the time-path of crimes covered by the sentence enhancement to that
of uncovered crimes. The earliest literature (Loftin and McDowall 1981; Loftin,
Heumann and McDowall 1983; Loftin and McDowall 1984; McDowall, Loftin
and Wiersma 1992) considered the effects of sentence enhancements for specific
29 A
2009 review of the literature by Donohue reaches a similar conclusion.
20
crimes — particularly gun crimes, generally finding little evidence in favor of
deterrence. A more recent paper studies the impact of changes in sentencing
in the aftermath of London’s 2011 riots. Leveraging the fact that judges in the
U.K. handed down harsher sentences for “riot offenses” in the six months following the riots, Bell, Jaitman and Machin (2013) find evidence of sizeable declines
in riot offenses relative to non-riot offenses which, in the absence of changes in
policing, they attribute to the advent of a harsher sanctions regime. This claim
is bolstered by the fact that there was a relative decline in riot offenses in sectors
that experienced the brunt of the 2011 riots as well as sectors that saw no riot
activity.30
A second class of studies has examined the impact of changes in the sanctions
regime that have heterogeneous impacts on different groups of offenders. For
example, Drago, Galbiati and Vertova (2009) study the effect of a 2006 collective
clemency of incarcerated prisoners in Italy. Prisoners incarcerated prior to May
2006 were released from prison with the remainder of their sentences suspended
while prisoners incarcerated after May 2006 were ineligible for the clemency.
Released prisoners, however, were subject to a sentence enhancement for any
future crimes committed that were serious enough to merit a sentence of at least
two years. For such crimes, the sentence would be augmented by adding the
amount of time the prisoner was sentenced to serve prior to his pardon to his new
sentence. Thus the intervention created a situation in which otherwise similar
individuals convicted of the same crime faced dramatically different sanctions
regimes. The results of this natural experiment suggest an elasticity of crime
with respect to sentence length of approximately -0.5 at one year follow up.
Utilizing the same natural experiment, Buonanno and Raphael (2013) report
evidence that incapacitation effects forgone as a result of the collective clemency
were large, thus constraining the magnitude of the deterrence effect.
Similar findings are reported for the United States by Helland and Tabarrok
(2007). Using data from Californias three strikes regime, Helland and Tabarrok (2007) compare the criminal behavior of individuals convicted of a second
“strikeable” offense to those tried for a second strikeable offense but who were
ultimately convicted of a lesser offense. The authors find evidence of an appreciable deterrent effect, calculating that California’s three strikes legislation
reduced felony arrest rates by approximately 20 percent among criminals with
two strikeable offenses against them on their record. Similarly while Zimring,
Hawkins and Kamin (2001) find little evidence of an overall effect of three strikes
legislation, they do find evidence that individual offenders with two strikes are
less likely to be arrested. Given that the deterrence margin is most salient
at two strikes, these studies stand out as especially important with respect to
identifying a meaningful deterrence effect of sentencing. On the other hand,
the magnitude of the response is actually quite small once one considers the
increase in sentence lengths associated with three strikes. Helland and Tabarrok’s estimates suggest to an elasticity of crime with respect to sentence length
30 Sentencing did change along both the intensive and extensive margins indicating the
incapacitation cannot be ruled out.
21
of -0.06.
3.2
Capital Punishment Regimes
Variation in the presence or intensity of capital punishment generates a potentially excellent source of variation with which to test for the magnitude of
general deterrence. In particular, to the extent that variation in a state’s capital punishment regime is unrelated to changes in the intensity of policing, the
effect of capital punishment represents a pure measure of deterrence with any
response of murder to the presence or intensity of capital punishment not plausibly attributable to incapacitation.31
There are been two primary approaches to identifying deterrence effects
of capital punishment. One approach considers the use of granular time series
data or event studies to identify the effect of the timing of executions on murder.
Time series studies typically use vector autoregression to assess whether murder
rates appear to decline in the immediate aftermath of an execution. Prominent
examples include Stolzenberg and D’Alessio (2004) which finds no evidence of
deterrence and Land, Teske and Zhang (2009) which finds evidence of short-run
deterrence. Event studies such as those of Grogger (1990) and Hjalmarrson
(2009) examine the daily incidence of homicides before and after executions.
Both Grogger (1990) and Hjalmarrson (2009) find little evidence of deterrence
effects though as Charles and Durlauf (2012) and Hjalmarrson (2012) note, with
a limited time horizon, it is not possible to distinguish between deterrence and
temporal displacement. A related study, Cochran, Chamblin and Seth (1994)
considers the effect of Oklahoma’s first execution in more than twenty years
and finds evidence that the execution appears to have increased murder among
strangers, an effect they attribute to a “brutalization” hypothesis though is
equally well attributed to statistical noise. A final study worth noting is that
of Zimring, Fagan and Johnson (2010) who compare homicide rates between
Singapore which uses the death penalty with variable intensity and Hong Kong
which does not use the death penalty. The paper finds no evidence in favor of
deterrence as both countries experience similar homicide trends over the thirtyfive year time period studied.
Broadly speaking, the time series and event studies literatures offer little
support in favor of deterrence though, as noted by Charles and Durlauf (2012),
the literature is plagued by several conceptual problems which compromise the
interpretability of estimated treatment effects. In particular, the focus of the
time series literature on executions as opposed to the sanctions regime more
generally marks a divergence from the neoclassical model of crime insofar as the
occurence of an execution does not per se change the expected severity of a criminal sanction for murder.32 Indeed the research design is often motivated by the
31 The argument is that in the absence of a capital punishment regime or a death sentence,
a convicted offender would nevertheless be sentenced to a lengthy prison sentence such as a
life sentence without the possibility of parole.
32 An important exception to this general point can be found in Chen and Horton (2012)
which studies the effect of executions for desertion among British soldiers during World War
22
assumption that an execution affects an offender’s perceived sanction. However,
there is little evidence, empirical or otherwise to support this assumption. Second, Charles and Durlauf note that the underlying logic of time series analyses
of executions and murder operationalize as deterrence the dynamic correlations
between a shock to one time series and the levels of another. As the authors
note, this is an arbitrary conceptualization of what is meant by deterrence.
A second literature studies the deterrent effect of capital punishment utilizing panel data on U.S. states to identify the effect of a capital punishment
statute or the frequency of executions on murder among the public at large. In
particular, these studies have exploited the fact that in addition to cross-state
differences in sentencing policy, there is also variation over time for individual states in the official sentencing regime, in the propensity to seek the death
penalty in practice, and in the application of the ultimate punishment (Chalfin,
Haviland and Raphael 2012). This literature has generated mixed findings with
several prominent papers (e.g., Dezbakhsh, Rubin and Shepherd 2003; Mocan
and Gittings 2003; Zimmerman 2004, 2006; Dezbakhsh and Shepherd 2006) finding large and significant deterrence effects and several equally prominent papers
(Katz, Levitt and Shustorovich 2003; Berk 2005; Donohue and Wolfers 2005,
2009; Kovandzic, Vieraitis and Paquette-Boots 2009) finding little evidence in
favor of deterrence.33
While evidence in favor of deterrence is mixed, recent reviews by Donohue
and Wolfers (2005, 2009) and Chalfin, Haviland and Raphael (2012) as well as
a 2011 report commissioned by the National Academy of Sciences point to substantial problems in a number of papers that purport to find deterrence effects
of capital punishment. These problems include the use of weak and/or inappropriate instruments (Dekbakhsh, Rubin and Shepherd 2003; Zimmerman 2004),
failure to report standard errors that are robust to within-state dependence
(Dezbakhsh and Shepherd 2006; Zimmerman 2009) and sensitivity of estimates
to different conceptions of perceived execution risk (Mocan and Gittings 2003).34
More generally, the panel data literature suffers from the threat of policy endogeneity, failure to include accurate controls and a lack of knowledge regarding
how potential offenders perceive execution risk. Finally, as noted by Berk (2005)
and Donohue and Wolfers (2005), results are highly sensitive to the inclusion of
certain states and even certain influential data points (i.e., Texas in 1997). The
most careful paper to date is that of Kovandzic, Vieritis and Paquette-Boots
(2009) who use a dataset spanning a longer period of time, employ an expanded
set of control variables and explore a wide variety of operationalizations of the
I and finds evidence that executions deter desertion, but actually encouraged desertion when
the execution was for an offense other than desertion or if the executed soldier was Irish. The
reason why this study stands as an exception to the rule proposed by Charles and Durlauf is
that during a time of war, the sanction regime is likely to be in constant flux.
33 The debate continues with recent responses to critiques by Donohue and Wolfers (2005)
offered by Zimmerman (2009), Dezbakhsh and Rubin (2010) and Mocan and Gittings (2010).
34 While Mocan and Gittings (2010) provide an extensive summary of the robustness of
results reported in Mocan and Gittings (2003), Chalfin, Haviland and Raphael (2012) point out
that the responsiveness of murder to execution risk relies on the assumption that individuals
are executed fairly soon (within six years) of a conviction.
23
effect of capital punishment and execution risk. The authors find no evidence
of a deterrent effect.
3.3
Sanction Nonlinearities
A second literature that seeks to estimate the magnitude of deterrence effects
does so by exploiting nonlinearities in the severity of sanctions faced by certain
offenders. Typically, these studies estimate the incidence of arrest rates for
young offenders who are either just below or just above the age of criminal
majority — either 17 or 18 years of age depending on the state. While offenders
below a given state’s age cutoff are treated as juveniles, offenders who are just
above the age of majority are tried as adults and are subject to considerably
more severe sanctions. Given that the conditional probability of an arrest is
smooth as a function of age around the age of criminal majority, any behavioral
response of offenders to the threshold is likely to be due to deterrence.
Using data from Florida, Lee and McCrary (2009) document a sizeable discontinuity in the probability that a young offender is sentenced to prison depending upon whether the arrest occured prior to or after the offender’s eighteenth birthday. Despite the fact that the expected sentence length for an adult
arrestee is over twice as great as that faced by a juvenile offender, Lee and McCrary find little evidence of deterrence. Their estimates suggest an elasticity
of crime with respect to sentence lengths of approximately -0.05, an estimate
that is far smaller than that of Drago, Galbiati and Vertova (2009). Findings
in Lee and McCrary are perhaps surprising but are supported by results reported in Hjalmarrson (2009) who documents that perceived increases in the
severity of sanctions at the age of criminal majority among juvenile offenders
are smaller than the actual changes. The implication is that deterrence is not
operational because perceptions do not match the incentives created by public policy. Indeed, in a reduced form analysis using national-level data in the
National Longitudinal Survey of Youth (NLSY), Hjalmarrson (2009) finds little
evidence of deterrence using self-reported data on offending.
A final paper worth mentioning is that of Hjalmarrson (2008) which studies
the effect of serving time in prison on subsequent arrest among juvenile offenders in Washington State. Exploiting a discontinuity in the state’s sentencing
guidelines, Hjalmarrson reports that incarcerated juveniles have lower propensities to be reconvicted of a crime. This deterrent effect is also observed for
older and for more criminally experienced offenders. The differential findings
in Hjalmarrson (2008) on the one hand and Hjalmarrson (2009) and Lee and
McCrary (2009) on the other hand can potentially be rationalized by the fact
that while the latter studies considered general deterrence (the behavioral response to a general threat of punishment), the former study measures specific
deterrence — that is, the behavioral response to actual punishment that has
already been experienced.
24
3.4
Deterrence vs. Incapacitation
As with the literature examining the response of crime to the certainty of apprehension, the primary conceptual challenge to interpreting the empirical literature on sanctions is that it is difficult to discern between deterrence and
incapacitation. With respect to studies of the crime-prison population elasticity, two issues merit discussion. First, the size of a state’s prison population
is only a proxy for the punitiveness of the sanctions regime. In practice, the
size of the prison population is a function of many things: the underlying rate
of offending, the certainty of punishment (in part due to the probability of
apprehension) and the criminal propensity of the marginal offender when the
prison population changes. Prison population is a stock not a flow and accordingly when the prison population declines it can be due to either an increase in
the contemporary probability of a custodial sentence or to flows out of prison
(Durlauf and Nagin 2011). Likewise, deterrence is only one of the mechanisms
by which prisons affect crime, the other being incapacitation. For these reasons,
the literature that examines the crime-prison population elasticity while important with respect to public policy is not particularly informative with respect
to deterrence.
In our view, research that studies the instantaneous impact of shocks to
the sanctions regime are considerably more informative. Indeed identifying the
sensitivity of crime to a shock to the sanctions regime is conceptually close
to testing Becker’s prediction that behavior will respond to the severity of a
sanction. However, even with perfect identification, attributing a change in
offending that occurs in the aftermath of a sanctions shock to deterrence requires
a logical leap. In particular, the logical leap is greatest when the sanctions
regime becomes more punitive along both the intensive and extensive margin.
To the extent that a custodial sentence becomes both longer and more likely,
tougher sentencing generates both deterrence and incapacitation effects. This is
an issue in interpreting much of the literature on sentence enhancements. Such
a concern is addressed in Kessler and Levitt (1998) which studies the effect
of California Proposition 8, a ballot amendment that enhanced the length of
sentences for certain felonies but not for others. Because prior to Proposition 8,
each of the felonies already required mandatory prison time, any instantaneous
response of crime to Proportion 8 would have to be attributable to deterrence.
Kessler and Levitt find that crimes that were eligible for the enhancement fell
by between 4 and 8 percent in the aftermath of Proposition 8 relative to a
control group of crimes not eligible for the enhancement. The implication of
these findings is that increased sanctions promote substantial deterrence. The
validity of Kessler and Levitt’s results have been called into question by Webster,
Doob and Zimring (2006) who argue that crime did not fall in the aftermath
of Proposition 8 and by Raphael (2006) who argues that crimes ineligible for
sentence enhancements do not form an appropriate control group for crimes
eligible for the enhancement.
Changes in a state’s use of capital punishment, in theory, offers a more appropriate means of identifying deterrence. This is because capital murder is
25
sufficiently serious as to warrant prison time regardless of the specifics of the
sentencing regime. Hence, when an offender is sentenced to death (as opposed
to a sentence of life without the possibility of parole), there is no instantaneous
incapacitation effect. With respect to capital punishment, the evidence of deterrence is, at best, mixed with the most rigorous studies failing to find evidence
of deterrence. Moreover, the identification problems in the literature are considerable as it is difficult to identify a shock to a state’s capital punishment regime
that is plausibly exogenous.
Undoubtedly the best tests for deterrence may be found in research that
follows individual offenders who, upon being apprehended, face different sanctions for a given crime. To the extent that differential treatment is driven by
arbitrary distinctions within the criminal justice system, research can identify
deterrence by comparing the behavior of offenders who are otherwise similar
but are treated differently. Such a research design is truly quasi-experimental
in the sense that treatment effects can be interpreted using the language of the
Rubin causal model. Moreover, individual-level studies track the behavior of
individuals who are not in prison and accordingly are not incapacitated, any
behavioral shift is plausibly attributable to deterrence. These individual-level
studies produce mixed evidence. On the one hand, studies of three strikes laws
establish that offenders with two strikes are less likely to reoffend than offenders with one strike (Zimring, Hawkins and Kamin 2001; Helland and Tabbarok
2007). Likewise, in Drago, Galbiati and Vertova’s study of Italy’s clemency bill,
prisoners who faced harsher sanctions upon being re-arrested were less likely
to be re-arrested. On the other hand, studies of sanction non-linearities in
which offenders of slightly different ages receive differential treatment report
little evidence of a large deterrence effect. These results might be rationalized
by differences in the responsiveness to a sanction among offenders of different
ages.
To date, the degree to which offenders are deterred by harsher sanctions
remains an open question. Undoubtedly deterrence can exist in extreme circumstances in which the punishment is immediate and harsh. However, within
the range of typical changes to sanctions in contemporary criminal justice systems, the evidence suggests that the magnitude of deterrence is not large and
is likely to be smaller than the magnitude of deterrence induced by changes in
the certainty of capture.
4
Work and Crime
The final pillar of the neoclassical model of crime considers the responsiveness
of crime to a carrot (better employment opportunities) rather than a stick (certainty or severity of punishment). In particular, since the benefit of a criminal
act must be weighed against the value of the offender’s time spent in an alternative activity, an increase in the oportunity cost of an offenders’ time can be
thought of as a deterrent to crime. Indeed, this principle has generated considerable public support for a variety of policies designed to reduce recidivism
26
among offenders returning from prison — in particular, the provision of job
training, employment counseling and transitional jobs.
The empirical literature examining the impact of local labor market conditions on crime can be divided into two related but distinct research literatures.
The first literature examines the relationship between unemployment and crime.
A second literature examines the impact of responsiveness of crime to wages.35
With respect to both literatures, approaches to study the effect of labor markets
on crime are varied and include papers that use individual microdata as well as
state- or county-level variation. Taken as a whole, the literature that uses aggregate data to disentangle the effect of economic conditions on crime presents
a mixed picture. In general, results are sensitive to the time period studied, the
population under consideration, the type of wage or unemployment rate that is
employed, as well as the criminal offenses analyzed. However, more recent and
carefully identified papers tend to find evidence of a fairly robust relationship
between both unemployment and wages and crime.
4.1
Unemployment
Periods of unemployment are thought to generate incentives to engage in criminal activity either as a means of income supplementation or consumption smoothing or, more generally, due to the effect of psychological strain (Chalfin and
Raphael 2011). To the extent that a decline in unemployment raises the opportunity cost of crime without generating a subsequent increase in the probability
of apprehension or the severity of the expected sanction, the response of crime
to changes in the unemployment rate can be thought of as capturing, among a
host of behavioral responses, deterrence.
In general, the early literature linking unemployment and crime has produced mixed and frequently contradictory results leading Chiricos (1987) to
characterize scholarly opinion on the topic as a “consensus of doubt.” In particular, Chiricos found that, among the studies he reviewed, fewer than half found
significant positive effects of aggregate unemployment rates on crime rates.36
This conclusion is echoed in reviews by Freeman (1983), Piehl (1998), Mustard
(2010) and Chalfin and Raphael (2011).
Recent literature on the topic of unemployment and crime has benefited from
several methodological advances — in particular, the use of panel data as opposed to a cross-sectional data or national time series. Examples of panel data
35 There is also a large and growing experimental literature that evaluates how at-risk individuals have responded to the provision of job coaching, employment counseling, career
placement and other employment-based services.
36 Nonetheless, Chiricos review also found that the unemployment-crime relationship was
three times more likely to be positive than negative and fifteen times more likely to be positive
and significant than negative and significant, indicating a basis for further research. The
results were especially strong for property crimes; in particular, for larceny and burglary.
Chiricos suggests that research results are generally consistent by level of aggregation though
they tend to be more consistently positive and significant at lower levels of aggregation.
This hypothesis is echoed by Levitt (2001) who likewise argues that national-level time-series
analyses obscure the unemployment-crime relationship by failing to account for rich variation
across space.
27
research include Entorf and Spengler (2000) for Germany, Papps and Winkelmann (2002) for New Zealand, Machin and Meghir (2004) for the United Kingdom and Arvanites and Defina (2006) and Ihlanfeldt (2007) for the United
States. With the exception of Papps and Winkelmann (2002), each of these
papers finds evidence in favor of a link between unemployment and crime, in
particular, property crime.
A second innovation in the recent literature has been to employ instrumental
variables to address the potential endogeneity between labor market conditions
and crime. The first such study is that of Raphael and Winter-Ebmer (2001)
who use a state-level panel data set covering 1979-1998 to study the effect of
unemployment rates on various types of crime employing two instruments for
the unemployment rate – the value of military contracts with the federal government as well as the regional impact of shocks to the price of oil. For property
crime rates, the results consistently indicate a positive effect of unemployment
on crime with a one percentage point increase in the unemployment rate predicting a 3-5 percent increase in property crime. For violent crime, however,
the results are mixed. Gould, Weinberg and Mustard (2002) provide a similar
analysis at the county level, using a county’s initial industry mix and measures
of skill-biased technical change as an instrument for unemployment. They too
find evidence of a positive relationship between unemployment and crime, particularly property crime. Taken as a whole, results reported in Raphael and
Winter-Ebmer (2001) and Gould, Weinberg and Mustard (2002) imply that
variation in unemployment rates explained between 12 percent and 40 percent
of the decline in property crime during parts of the 1990s. In a more recent paper, Lin (2008) builds on these approaches using exchange rate shocks to isolate
exogenous variation in unemployment rates. Lin reports that a one percentage
point increase in unemployment leads to a 4 to 6 percent decline in property
crime and would explain roughly one third of the crime drop during the 1990s.37
On the whole, the preponderance of the evidence suggests that there is an
important relationship between unemployment rates and property crime but
little impact of unemployment on violent crime. In the recent literature which is
more careful with respect to addressing omitted variables bias and simultaneity,
the relationship between unemployment and property crime is found regardless
of the level of aggregation (counties or states).38 The relationship between
unemployment and property crime is empirically meaningful as property crime
would be predicted to rise by between 9 and 18 percent during a serious recession
37 Fougere
et al. (2009) provide a similar analysis for France, finding effects that are similar
in magnitude.
38 Prominent IV papers including Raphael and Winter-Ebmer (2001), Gould, Weinberg and
Mustard (2002) and Lin (2008) do not uniformly find that instrumening results in a more
positive relationship between unemployment and crime as would be predicted by the omission
of pro-cyclical control variables or simultaneity bias. Another explanation for slippage between
least squares and IV estimates of the effect of unemployment on crime is measurement errors
in the unemployment rate. To the extent that such errors are classical, attenuation bias
will mean that 2SLS estimates will exceed OLS estimates. To the extent that this pattern
is not found, there is the possibility that OLS estimates are actually upward biased due to
simultaneity or omitted variables.
28
in which unemployment increased by three percentage points. Moreover, this,
if anything, may understate the magnitude of the relationship as crime appears
to be particularly sensitive to the existence of employment opportunities for
low skilled-men (Schnepel 2013). Nevertheless, the estimates remain sensitive
to the time period studied. To wit, property crime has generally continued to
decline through the recent Great Recession which increased unemployment rates
nationally by as many as four percentage points.
4.2
Wages
A second and related research literature considers the impact of wage levels on
crime rates. There are several a priori reasons to expect a stronger relationship
between crime and wages than between crime and unemployment. First, as
noted by Gould, Weinberg and Mustard (2002), since criminal participation is
associated with a set of fixed costs, crime may well be more responsive to long
term labor market measures such as levels of human capital or wages than to
unemployment spells, which are typically ephemeral. Second, at any given time,
the number of individuals who are employed in low-wage jobs vastly outnumbers
the number of unemployed and, as such, wages for unskilled men may play a
proportionally greater role than unemployment in encouraging crime (Hansen
and Machin 1999). In fact, among individuals who reported engaging in crime
during the past year, a large majority reported wage earnings (Grogger 1998)
and three quarters were employed at the time of their arrest (Lynch and Sabol
2001), indicating that the behavior of a majority of offenders should be sensitive
to changes in the wage.
The literature linking wages to crime has, in general, generated more consensus than the unemployment literature. Prominent panel data papers include
Doyle, Ahmed and Horn (1999) who analyze state-level panel for 1984-1993 and
find that higher average wages reduce both property and violent crime (elasticity estimates vary between -0.3 and -0.9) and Gould, Weinberg and Mustard
(2002) who restrict their analysis to the wages of relatively low skilled men and
find, using a county-level panel spanning 1979 to 1997, that the falling wages
of unskilled men in this period led to an 18 percent increase in robbery, a 14
percent increase in burglary and a 7 percent increase in larceny. These findings
are striking in that they indicate that wage trends explain more than half of the
increase in both violent and property crimes over the entire period.
In a similar analysis for the U.K., Machin and Meghir (2004) examined
changes in regional crime rates in relation to changes in the 10th and 25th
percentile of the regions wage distribution and focus on the retail sector, an
industry where low skilled workers have the ability to manipulate their hours of
work. They find that crime rates are higher in areas where the bottom of the
wage distribution is low. With regard to microdata, Grogger (1998), leveraging
data from the NLSY , finds that youth wages account for approximately three
quarters of the variation in youth crime. Finally, a related literature considers
the responsiveness of crime to minimum wages and consistently finds evidence in
favor of a negative relationship between the two variables (Corman and Mocan
29
2005; Hansen and Machin 2006; Fernandez, Holman and Pepper 2012).
4.2.1
Identifying Deterrence
A first order issue in interpreting research on the effect of police and prisons
on crime concerns the extent to which deterrence can be disentangled from
incapacitation. This issue is not relevant in considering the responsiveness of
crime to changes in wages or employment conditions. Nevertheless, it is worth
considering whether a significant coefficient on the wage or the unemployment
rate in a crime regression necessarily identifies deterrence. In particular, for
several reasons, crime and unemployment may be related for reasons other than
deterrence. First, a relationship between macroeconomic conditions and crime
may exist due to the relationship between macroeconomic conditions and criminal opportunities (Cook 1985). For example, during a recession auto thefts
tend to decline presumably because fewer people are employed and therefore
drive their cars (Cook 2010). Second, employment conditions and crime may
be linked through behavioral changes that cannot be properly characterized as
deterrence. For example, a displaced worker may well have feelings of anger or
loss of control that manifest themselves in violent behavior. In such a case, the
job may not be protective against crime through any deterrence mechanism per
se. Nevertheless a robust relationship between economic conditions and crime
is potentially consistent with the idea that individuals respond to incentives, at
least at the margin.
5
Conclusion
We reviewed three large literatures regarding the responsiveness of crime to
police, sanctions and local labor market opportunities. Three key conclusions
are worth noting. First, there is robust evidence that crime responds to increases
in police manpower and to many varieties of police redeployments. With respect
to manpower, our best guess is that the elasticity of violent crime and property
crime with respect to police are approximately -0.4 and -0.2, respectively. The
degree to which these effects can be attributed to deterrence as opposed to
incapacitation remains an open question though analyses of arrest rates suggests
a role for deterrence (Levitt 1998, Owens 2013). With respect to deployments,
experimental research on hot spots policing and focused deterrence efforts have,
in some cases, led to remarkably large decreases in offending.
Second, while the evidence in favor of a crime-sanction link is generally
mixed, there does appear to be some evidence of deterrence effects induced by
policies that target specific offenders with sentence enhancements. This is seen
in the effect of California’s three strikes law on the behavior of offenders with
two strikes (see Helland and Tabarrok 2007) and in the behavior of pardoned
Italian offenders (Drago, Galbiati and Vertova 2009). On the other hand, while
the elasticity of crime with respect to sentence lengths appears to be large in
the Italian case, it is quite small in the California case.
30
Finally, there is fairly strong evidence, in general, of a link between local
labor market conditions, proxied using the unemployment rate or the wage, on
crime. While these effects are unlikely to be appreciably contaminated by incapacitation effects, they may reflect behavioral responses aside from deterrence.
Overall, the evidence suggests that individuals respond to the incentives
that are the most immediate and salient. While police and local labor market
conditions influences costs that are borne immediately, the cost of a prison sentence, if experienced at all, is experienced sometime in the future. To the extent
that offenders are myopic or have a high discount rate, deterrence effects will
be less likely. Moreover, given that an empirical finding of deterrence depends
on the existence of perceptual deterrence, it may be the case that potential offenders are more aware of changes in policing and local labor market conditions
than they are of changes in incarceration policy, with the exception of specific
sentence enhancements that are individually targeted.
In closing, we note that Gary Becker’s recent passing prompts us to acknowledge yet again the decisive impact of his landmark 1968 paper. As his ubiquity
in this review makes clear, his insights launched an entire literature—one that
has had and continues to have profound implications for and impact on public
policy and safety.
31
References
Allingham, Michael G. and Agnar Sandmo (1972). “Income Tax Evasion: A Theoretical
Analysis,” Journal of Public Economics, 1(3), 323-338.
Andenaes, Johannes (1974). Punishment and Deterrence. Ann Arbor: University of
Michigan Press.
Anwar, Shamena and Thomas Loughran (2011). “Testing a Bayesian Learning Theory of
Deterrence Among Serious Juvenile Offenders,” Criminology 49(3): 667-698.
Apel, Robert and Daniel S. Nagin (2010). “Deterrence,” in Crime, 4th Edition, James W.
Wilson and Joan Petersilia (eds.), Oxford University Press.
Arvanites, Thomas M. and Robert H. Defina (2006). “Business Cycles and Street Crime,”
Criminology, 44(1), 139-164.
Ayres, Ian and Steven D. Levitt (1998). “Measuring Positive Externalities from
Unobservable Victim Precaution: An Empirical Analysis of Lojack,” The Quarterly
Journal of Economics 113 (1): 43–77.
Baldry, Jonathan (1974), “Positive Economic Analysis of Criminal Behaviour,”
Economic Policies and Social Goals, 171-198.
Beccaria, Cesare (1764). On Crimes and Punishments, Oxford: Clarendon Press.
Becker, Gary (1968). “Crime and Punishment: an Economic Analysis.” Journal of
Political Economy 78, 2, 169-217.
Bell, Brian, Laura Jaitman and Stephen Machin (2014). “Crime Deterrence: Evidence
From the London 2011 Riots,” forthcoming in Economic Journal.
Benson, Bruce L. and Brent D. Mast (2001). “Privately Produced General Deterrence,”
Journal of Law and Economics. 44 (2): 725-746.
Bentham, Jeremy (1798). An Introduction to the Principles of Morals and Legislation,
Oxford: Clarendon Press.
Berk, Richard (2005). “New claims about executions and general deterrence: deja vu all
over again?” Journal of Empirical Legal Studies 2:303–330.
Berk, Richard and John MacDonald (2010). “Policing the Homeless,” Criminology &
Public Policy, 9(4), 813-840.
Block, M. K., & Heineke, J. M. (1975). “A Labor Theoretic Analysis of the Criminal
Choice,” American Economic Review, 65(3), 314-325.
Braga, Anthony A. 2008. Police Enforcement Strategies to Prevent Crime in Hot Spot
Areas. Crime Prevention Research Review (No. 2). Washington, DC: Office of
Community Oriented Policing, U.S. Department of Justice.
Braga, Anthony A., David L. Weisburd, Elin G. Waring, Lorraine Mazerolle, William
and Francis Gajewski, (1999). “Problem-Oriented Policing in Violent Crime Places: A
Randomized Controlled Experiment,” Criminology, 37(3), 541-580.
Braga, Anthony A., “The Effects of Hot Spots Policing on Crime,” The Annals of the
American Academy of Political and Social Science, 2001, 578, 104–125.
Braga, Anthony A., “Hot Spots Policing and Crime Prevention: A Systematic Review of
Randomized Controlled Trials,” Journal of Experimental Criminology, September 2005,
1 (3), 317–342.
Braga, Anthony A. (2008). Police Enforcement Strategies to Prevent Crime in Hot Spot
Areas. Crime Prevention Research Review (No. 2). Washington, DC: Office of
Community Oriented Policing, U.S. Department of Justice.
Braga, Anthony A. and Brenda J. Bond, “Policing Crime and Disorder Hot Spots: A
Randomized Controlled Trial,” Criminology, 2008, 46 (3), 577–607.
Braga, Anthony A., David M. Kennedy, Elin J. Waring and Anne Morrison Piehl (2001).
“Problem-oriented policing, deterrence, and youth violence: An evaluation of Boston's
Operation Ceasefire,” Journal of Research in Crime and Delinquency, 38(3), 195-225.
Buonanno, Paolo, & Steven Raphael (2013). “Incarceration and Incapacitation: Evidence
from the 2006 Italian Collective Pardon,” The American Economic Review, 103(6), 24372465.
Burdett, Kenneth, Ricardo Lagos and Randall Wright (2004). “An On-the-Job Search
Model of Crime, Inequality and Unemployment,” International Economic Review, 45(3),
681-706.
Caetano, Gregorio and Vikram Maheshri (2014). “Identifying Dynamic Spillovers in
Criminal Behavior,” Working Paper.
Caeti, T. (1999). Houston’s targeted beat program: A quasi-experimental test of police
patrol strategies. Ph.D. dissertation, Sam Houston State University. Ann Arbor, MI:
University Microfilms International.
Cameron, Samuel (1988). The Economics of Crime Deterrence: A Survey of Theory and
Evidence. Kyklos, 41(2), 301-323.
Chalfin, Aaron and Justin McCrary (2013). “Are U.S. Cities Underpoliced? Theory and
Evidence,” NBER Working Paper #18815.
Chalfin, Aaron and Steven Raphael (2011). “Work and Crime,” Oxford Handbook on
Crime and Criminal Justice, Michael Tonry, ed.
Chalfin, Aaron, Amelia M. Haviland, and Steven Raphael (2013). "What do Panel
Studies Tell Us about a Deterrent Effect of Capital Punishment? A Critique of the
Literature." Journal of Quantitative Criminology 29(1): 5-43.
Charles, Kerwin and Steven Durlauf (2013). “Pitfalls in the Use of Time Series Methods
to Study Deterrence and Capital Punishment,” Journal of Quantitative Criminology,
29(1), 45-66.
Chen, Daniel L., and John J. Horton (2012). "The Deterrent Effect of the Death Penalty?
Evidence from British Commutations During World War I,” Working Paper.
Chiricos, Theodore (1987). “Rates of Crime and Unemployment: An Analysis of
Aggregate Research Evidence.” Social Problems 34(2): 187–211.
Cochran, John K., Mitchell B. Chamblin., and Mark Seth (1994). Deterrence or
Brutalization? An Impact Assessment of Oklahoma’s Return to Capital Punishment,”
Criminology, 32(1), 107-134.
Cohen, Jacqueline and Jens Ludwig (2003). “Policing Gun Crime.” In Evaluating Gun
Policy, J. Ludwig and P. Cook, eds. Washington DC: Brookings Institution Press.
Cook, Philip J. (1980). “Research in Criminal Deterrence: Laying the Groundwork for
the Second Decade.” Crime and Justice 2: 211-268.
Cook, Philip J. (2010). “Property Crime—Yes; Violence—No,” Criminology & Public
Policy, 9(4), 693-697.
Cook, Philip J., and John M. MacDonald (2011). “Public Safety through Private Action:
an Economic Assessment of BIDS,” Economic Journal, 121(552), 445-462.
Cook, Philip J., and Gary A. Zarkin (1985). “Crime and the Business Cycle.” Journal of
Legal Studies 14(1): 115–28.
Corman, Hope and Naci Mocan (2005). “Carrots, Sticks, and Broken Windows.” Journal
of Law and Economics 48: 235-266.
Corsaro, Nicholas, Rod Brunson and Edmund McGarrell (2010). “Evaluating a Policing
Strategy Intended to Disrupt an Illicit Street-Level Drug Market. Evaluation Review,
34(6), 513-548.
Corsaro, Nicholas, Aaron Chalfin and Edmund McGarrell (2014). “An Application of
Synthetic Controls for Comparative Case Studies: Examining the Impact of the National
Project Safe Neighborhoods Strategy on Urban Gun Violence,” Working Paper.
Corsaro, Nicholas and Edmund McGarrell (2010). “Reducing Homicide Risk in
Indianapolis Between 1997 and 2000,” Journal of Urban Health, 87(5), 851-864.
Corsaro N., Hunt E.D., Hipple N.K. and McGarrell E.F. (2012). “The impact of drug
market puling levers policing on neighborhood violence: an evaluation of the High Point
drug market intervention,” Criminology and Public Policy 11: 165-199
D’Aneglo, Gregory and Benjamin Hansen (2013). “Life and Death in the Fast Lane:
Police Enforcement and Traffic Fatalities,” forthcoming in American Economic Review:
Economic Policy.
Dezhbakhsh, Hahem and Paul Rubin (2007). "From the “Econometrics of Capital
Punishment” to the “Capital Punishment” of Econometrics: On the Use and Abuse of
Sensitivity Analysis," Unpublished manuscript. Department of Economics, University of
Georgia.
Dezhbakhsh, Hashem, Paul Rubin, and Joanna Shepherd (2003). “Does Capital
Punishment Have a Deterrent Effect? New Evidence from Post-Moratorium Panel Data.”
American Law and Economics Review 5: 344-76.
Di Tella, Rafael and Ernesto Schargrodsky (2004). “Do Police Reduce Crime? Estimates
Using the Allocation of Police Forces after a Terrorist Attack.” American Economic
Review 94: 115-133.
Donohue, John J. and Justin Wolfers (2005). “Uses and Abuses of Empirical Evidence in
the Death Penalty Debate.” Stanford Law Review 58: 791-846.
Donohue, John J and Justin Wolfers. (2009). “Estimating the impact of the death penalty
on murder,” American Law and Economics Review, 11(2), 249-309.
Donohue, John J., Daniel E. Ho and Patrick Leahy. (2013). “Do police reduce crime?,”
Empirical Legal Analysis: Assessing the Performance of Legal Institutions, 125.
Doyle, Joanne M., Ehsan Ahmed, and Robert N. Horn (1999). “The Effects of Labor
Markets and Income Inequality on Crime: Evidence from Panel Data.” Southern
Economic Journal 65:717–31.
Draca, Mirko, Stephen Machin and Robert Witt (2011). “Panic on the Streets of London:
Police, Crime, and the July 2005 Terror Attacks,” The American Economic Review,
101(5), 2157-2181.
Drago, Francesco, Roberto Galbiati, and Pietro Vertova (2009). “The Deterrent Effects of
Prison: Evidence from a Natural Experiment.” Journal of Political Economy 117: 257280.
Durlauf, Steven N. and Daniel S. Nagin (2011). “Imprisonment and Crime,” Criminology
& Public Policy, 10(1), 13-54.
Eck, John E and Edward Maguire. (2000). “Have changes in policing reduced violent
crime? An assessment of the evidence, The Crime Drop in America.
Eck, John E. and David Weisburd.(1995). Crime Places in Crime Theory. In John E. Eck
and David Weisburd (Eds.), Crime and Place: Crime Prevention Studies, 4 (pp. 1-33).
Monsey, NY: Willow Tree Press.
Ehrlich, Isaac (1973). “Participation in Illegitimate Activities: A Theoretical and
Empirical Investigation.” Journal of Political Economy 81: 521-65.
Engel Robin S., Marie S. Tillyer and Nicholas Corsaro (2013). “Reducing gang violence
using focused deterrence: evaluating the Cincinnati initiative to reduce violence (CIRV),”
Justice Quarterly 30:403-439
Entorf, Horst, and Hannes Spengler (2000). “Criminality, social cohesion and economic
performance.” Wurzburg Economic Papers No. 22. Department of Economics, University
of Wurzburg, Wurzburg, Germany.
Evans, William and Emily Owens (2007). “COPS and Crime.” Journal of Public
Economics 91: 181-201.
Fagan, Jeffrey, Amanda Gellar, Garth Davies &and Valerie West (2009). “Street Stops
and Broken Windows Revisited: The Demography and Logic of Proactive Policing in a
Safe and Changing City,” Race, Ethnicity and Policing: New and Essential Readings,
Stephen K. Rice and Michael D. White, eds., New York University Press.
Fernandez, Jose M., Thomas Holman and John V. Pepper (2014). “The Impact of Living
Wage Ordinances on Crime,” forthcoming in Industrial Relations.
Fougère, D., Kramarz, F., and Pouget, J. (2009). “Youth unemployment and crime in
France,” Journal of the European Economic Association, 7(5), 909-938.
Freeman, Richard (1983). “Crime and unemployment,” Crime and Public Policy.
Gould, Eric D., Weinberg, Bruce A., and David B. Mustard (2002). “Crime Rates and
Local Labor Market Opportunities in the United States: 1979–1997,” Review of
Economics and Statistics 84(1): 45–61.
Greenwood, Peter and Angela Hawken (2002). “An Assessment of the Effect of
California’s Three- Strikes Law.” Greenwood Associates. Working Paper.
Griliches, Zvi and Jerry A. Hausman (1986). “Errors in Variables in Panel Data,” Journal
of Econometrics, 31(1), 93-118.
Grogger, Jeffrey (1991). “Certainty vs. Severity in Punishment.” Economic Inquiry 29(2):
297–309.
Grogger, Jeffrey (1995). “The Effect of Arrest on the Employment and Earnings of
Young Men.” Quarterly Journal of Economics 110(1)1: 51–71.
Grogger, Jeffrey (1998). “Market Wages and Youth Crime.” Journal of Labor Economics
16(4): 756–91.
Gronau, Reuben (1980). “Home Production--a Forgotten Industry,” The Review of
Economics and Statistics, 408-416.
Hansen, Kristine and Stephen Machin (2002). “Spatial Crime Patterns and the
Introduction of the UK Minimum Wage.” Oxford Bulletin of Economics and Statistics, 64
(August): 677-697.
Harcourt, Bernard E. and Jens Ludwig (2006). “Broken windows: New evidence from
New York City and a five-city social experiment,” The University of Chicago Law
Review, 271-320.
Helland, Eric and Alexander Tabarrok (2007). “Does Three Strikes Deter? A
Nonparametric Estimation,” Journal of Human Resources 42: 309-330.
Hjalmarsson, Randi (2009). “Crime and Expected Punishment: Changes in Perceptions at
the Age of Criminal Majority,” American Law and Economics Eeview, 11(1), 209-248.
Hjalmarsson, Randi (2009). “Juvenile Jails: A Path to the Straight and Narrow or to
Hardened Criminality?,” Journal of Law and Economics, 52(4), 779-809.
Randi Hjalmarsson (2012) “Can Executions Have a Short-Term Deterrence Effect on
Non-Felony Homicides?,” Criminology & Public Policy. 11(3): 565-571.
Huang, Chien-Chieh, Derek Laing, and Ping Wang (2004). “Crime and Poverty: A
Search- Theoretic Approach.” International Economic Review, 45 (3): 909-938.
Ihlanfeldt, Keith (2007). “Neighborhood Drug Crime and Young Males’ Job
Accessibility,” The Review of Economics and Statistics, 89 (1) (February): 151-164.
Imai, Susumu and Kala Krishna (2004). “Employment, Deterrence, and Crime in a
Dynamic Model.” International Economic Review 45: 845-872.
Johnson, Rucker and Steven Raphael (2012). “How Much Crime Reduction Does the
Marginal Prisoner Buy?,” Journal of Law and Economics, 55(2), 275-310.
Katz Lawrence, Steven D. Levitt and Ellen Shustorovich (2003). “Prison Conditions,
Capital Punishment, and Deterrence,” American Law and Economics Review 5(2): 318–
343.
Kelling, George L. and William H. Sousa (2001). Do Police Matter?: An Analysis of the
Impact of New York City's Police Reforms. CCI Center for Civic Innovation at the
Manhattan Institute.
Kennedy, David, Anthony A. Braga, Anne Morrison Piehl, and Elin J. Waring (2001).
Reducing Gun Violence: The Boston Gun Project’s Operation Ceasefire. Washington
DC: National Institute of Justice.
Kessler, Daniel and Steven D. Levitt (1999). “Using Sentence Enhancements to
Distinguish between Deterrence and Incapacitation,” Journal of Law and Economics 42:
343–63.
Klick, Jonathan and Alexander Tabarrok (2005). “Using Terror Alert Levels to Estimate
the Effect of Police on Crime.” Journal of Law and Economics 48: 267-279.
Kovandzic, Tomislav V., Lynne M. Vieraitis and Denise Paquette Boots (2009). “Does
the death penalty save lives?,” Criminology and Public Policy 8(4):803–843.
Kubrin, Charis E., Steven F. Messner, S. F., Glenn Deane, Kelly McGeever and Thomas
D. Stucky. (2010). “Proactive Policing and Robbery Rates Across U.S. Cities,”
Criminology, 48(1), 57-97.
Land, Kenneth C., Raymond H. Teske and Hui Zheng (2009). “The Short-Term Effects
of Executions on Homicides: Deterrence, Displacement or Both?,” Criminology, 47(4),
1009-1043.
Lee, David and Justin McCrary (2009). “The Deterrent Effect of Prison: Dynamic Theory
and Evidence,” Working Paper.
Lemieux, Thomas, Bernard Fortin. And Pierre Frechette (1994). “The Effect of Taxes on
Labor Supply in the Underground Economy. The American Economic Review, 84(1):
231-254.
Levitt, Steven D. (1996). The effect of prison population size on crime rates: Evidence
from prison overcrowding legislation. Quarterly Journal of Economics, 111: 319–352.
Levitt, Steven D. (1997). “Using Electoral Cycles in Police Hiring to Estimate the Effect
of Police on Crime.” American Economic Review 87: 270-290.
Levitt, Steven D. (1998). “Juvenile Crime and Punishment.” Journal of Political
Economy, 106: 1156- 1185.
Levitt, Steven D. (2002). “Using Electoral Cycles in Police Hiring to Estimate the Effect
of Police on Crime: Reply.” American Economic Review 92: 1244-1250.
Levitt, Steven D. (2006). “The Case of the Critics Who Missed the Point: A Reply to
Webster et. al.” Criminology and Public Policy 5: 449-460.
Levitt, Steven D., & Miles, Thomas J. (2006). “Economic contributions to the
understanding of crime,” Annual Review of Law and Social Science 2, 147-164.
Liedka, Raymond, Anne Morrison Piehl, and Bert Useem. 2006. The crime-control effect
of incarceration: Does scale matter? Criminology and Public Policy, 5: 245–276.
Lin, Ming-Jen (2008). “Does Unemployment Increase Crime? Evidence from the U.S.
Data 1974- 2000.” Journal of Human Resources, 43 (2): 413-436.
Lin, Ming-Jen (2009). “More Police, Less Crime: Evidence from U.S. State Data,”
International Review of Law and Economics, 29 (2), 73–80.
Lochner, Lance (2007). “Individual Perceptions of the Criminal Justice System,” The
American Economic Review 97(1): 444-460.
Lochner, Lance and Enrico Moretti (2004). “The Effect of Education on Crime: Evidence
from Prison Inmates, Arrests, and Self-reports.” The American Economic Review, 94 (1):
155-189.
Loftin, Colin, Milton Heumann, and David McDowall. (1983). “Mandatory Sentencing
and Firearms Violence: Evaluating an Alternative to Gun Control.” Law and Society
Review 17: 287–318.
Loftin, Colin and David McDowall (1981). ‘”’One with a Gun Gets You Two’:
Mandatory Sentencing and Firearms Violence in Detroit.” Annals of the American
Academy of Political and Social Science 455: 150–67.
Loftin, Colin and David McDowall (1984). “The Deterrent Effects of the Florida Felony
Firearm Law.”Journal of Criminal Law and Criminology 75: 250–59.
Lofstrom, Magnus and Steven Raphael (2013). “Public Safety Realignment and Crime
Rates in California,” Public Policy Institute of California.
MacDonald, John M. (2002). “The effectiveness of community policing in reducing
urban violence,” Crime & Delinquency, 48(4), 592-618.
MacDonald, John M., Jonathan Klick and Ben Grunwald (2013). “The Effect of Privately
Provided Police Services on Crime,” University of Pennsylvania Working Paper.
Machin, Stephen and Olivier Marie (2011). Crime and police resources: The street crime
initiative. Journal of the European Economic Association, 9(4), 678-701.
Machin, Stephen and Costas Meghir (2004). “Crime and Economic Incentives.” Journal
of Human Resources, 39 (4): 958-979.
Marvell, T. and C. Moody (1994). “Prison Population Growth and Crime Reduction.”
Journal of Quantitative Criminology 10: 109-140.
Marvell, Thomas B., & Moody, Carlisle E. (1996). Specification Problems, Police
Levels, And Crime Rates*. Criminology, 34(4), 609-646.
Maltz Michael D. and Joseph Targonski (2002). “A Note on the Use of County-Level
UCR Data,” Journal of Quantiative Criminology 18(3):297–318.
Mastrobuoni, Giovanni (2013). “Police and Clearance Rates: Evidence from Recurrent
Redeployments Within a City,” Working Paper.
McCrary, Justin. (2002). “Using Electoral Cycles in Police Hiring to Estimate the Effect
of Police on Crime: Comment.” American Economic Review 92: 1236-1243.
McCrary, Justin (2010). “Dynamic Perspectives on Crime,” in Bruce Benson and Paul R.
Zimmerman, eds., Handbook of the Economics of Crime, Northampton, MA: Edward
Elgar.
McDowall, David, Colin Loftin, and B. Wiersema (1992). “A Comparative Study of the
Preventive Effects of Mandatory Sentencing Laws for Gun Crime.” Journal of Criminal
Law and Criminology 83: 378-394.
McGarrell Edmund F. Cherman S, Weiss A, Wilson JM (2001). “Reducing firearms
violence through directed police patrol,” Criminology and Public Policy 1: 119-148
McGarrell EF, Chermak S, Wilson JM, Corsaro N (2006). “Reducing homicide through a
‘lever-pulling’ strategy,” Justice Quarterly 23:214–231
McGarrell EF, Corsaro N, Hipple NK, Bynum TS (2010). “Project safe neighborhoods
and violent crime trends in U.S. cities: Assessing violent crime impact,” Journal of
Quantitative Criminology 26: 165-190
McGarrell EF, Hipple NK, Corsaro N, Bynum TS, Perez H, Zimmerman CA, and Garmo,
M (2009). “Project Safe Neighborhoods – A National Program to Reduce Gun Crime:
Final Project Report. Washington DC: National Institute of Justice.
Mocan, Naci. and Kaj Gittings. (2003). “Getting Off Death Row: Commuted Sentences
and the Deterrent Effect of Capital Punishment,” Journal of Law and Economics
46(4):453–478.
Mocan, Naci. and Kaj Gittings (2010). “The Impact of Incentives on Human Behavior:
Can We Make it Disappear?: The Case of the Death Penalty,” in: Di Tella R, Edwards S,
Schargrodsky E (eds) The Economics of Crime; Lessons for and from Latin America.
University of Chicago Press.
Mustard, David B. (2010). “How Do Labor Markets Affect Crime? New Evidence on an
Old Puzzle,” Discussion paper series//Forschungsinstitut zur Zukunft der Arbeit.
Nagin, Daniel S. (1998). “Criminal Deterrence Research at the Outset of the Twenty-First
Century.” In Crime and Justice: A Review of Research, vol. 23, M. Tonry, ed. Chicago:
University of Chicago Press.
Nagin, Daniel S. (2013). “Deterrence: A Review of the Evidence by a Criminologist for
Economists,” Annual Review of Economics 5.
Nagin, Daniel S., Francis Cullen and Cheryl Lero Jonson (2009). “Imprisonment and ReOffending,” in Crime and Justice: A Review of Research 38, Michael Tonry (ed.),
University of Chicago Press.
Nagin, Daniel S. and Raymond Paternoster (1994). “Personal Capital and Social Control:
The Deterrence Implications of a Theory of Individual Differences in Criminal
Offending,” Criminology 32(4): 581-606.
Nagin, Daniel S. and Greg Pogarsky (2001). “Integrating Celerity, Impulsivity and
Extralegal Sanction Threats Into a Model of General Deterrence: Theory and Evidence,”
Criminology 39(4): 865-892.
Nagin, Daniel S. and Greg Pogarsky (2003). “Cheating as Crime: An Experimental
Investigation of Deterrence.” Criminology, 41: 167-194.
Nagin, Daniel S. and G. Matthew Snodgrass (2013). “The Effect of Incarceration on ReOffending: Evidence from a Natural Experiment in Pennsylvania,” Journal of
Quantitative Criminology
Owens, Emily G. (2013). “COPS and Cuffs,” Lessons from the Economics of Crime:
What Reduces Offending?, MIT Press.
Papachristos, Andrew V., Tracey Meares and Jeffrey Fagan (2007). “Attention Felons:
Evaluating Project Safe Neighborhoods in Chicago,” Journal of Empirical Legal Studies,
4(2), 223-272.
Papps, Kerry and Rainer Winkelmann (2002). “Unemployment and Crime: New
Evidence for an Old Question,” New Zealand Economic Papers, 34 (1): 53-72.
Petersilia, Joan and Elizabeth Piper Deschenes (1994). “Perceptions of Punishment:
Inmates and Staff Rank the Severity of Prison Versus Intermediate Sanctions,” The
Prison Journal, 74(3), 306-328.
Piehl, Anne Morrison. (1998). “Economics Conditions, Work and Crime,” in Handbook
of Crime and Punishment, Michel Tonry (ed.) New York: Oxford University Press.
Polinsky, A. Mitchell and Steven Shavell (1999). “On the Disutility and Discounting of
Imprisonment and the Theory of Deterrence.” Journal of Legal Studies XXVIII: 1-16.
Raphael, Steven (2006). “The Deterrent Effects of California’s Proposition 8: Weighing
the Evidence,” Criminology and Public Policy 5: 471-478.
Raphael, Steven and Jens Ludwig (2003). “Prison Sentence Enhancements: The Case of
Project Exile.” In Evaluating Gun Policy: Effects on Crime and Violence, Jens Ludwig
and Philip Cook, eds. Washington DC: Brookings Institution Press.
Raphael, Steven, and Rudolf Winter-Ebmer (2001). “Identifying the Effect of
Unemployment on Crime,” Journal of Law and Economics 44, no. 1: 259–84.
Roman, Caterina G. and Shannon Reid. 2012. “Assessing the Relationship between
Alcohol Outlets and Domestic Violence: Routine Activities and the Neighborhood
Environment,” Violence and Victims, 27(5): 811-828.
Roman, John and Aaron Chalfin (2007). “Has Demand for Crime Increased? The
Prevalence of Personal Media Devices and the Robbery Spike in 2005 and 2006,”
American Criminal Law Review 45: 1149-1166.
Rosenfeld, Richard, Robert Fornango and Eric Baumer (2005). “Did Ceasefire,
Compstat, and Excile Reduce Homicide?,” Criminology & Public Policy, 4(3), 419-449.
Rosenfeld, Richard, Robert Fornango and Andres Rengifo (2007). “The impact of order‐
maintenance policing on new york city homicide and robbery rates: 1988‐2001,”
Criminology, 45(2), 355-384.
Sampson, Robert J. and Jacqueline Cohen (1988). “Deterrent effects of the police on
crime: A replication and theoretical extension,” Law and Society Review, 163-189.
Schnepel, Kevin (2013). “Labor Market Opportunities and Crime: Evidence from
Parolees,” Working Paper.
Shaw, Clifford R. and Henry D. McKay (1942). Juvenile delinquency and urban areas.
Chicago, Ill.
Shepherd, Joanna (2002). Fear of the First Strike: The Full Deterrent Effect of
California’s Two‐ and Three‐Strikes Legislation, Journal of Legal Studies 31: 159-201.
Sherman, Lawrence W. and David Weisburd, “General Deterrent Effcts of Police Patrol
in Crime “Hot Spots”: A Randomized Controlled Trial,” Justice Quarterly, 1995, 12 (4),
625–647.
Sherman, Lawrence W. and Rogan, D. P. (1995). “Effects of Gun Seizures on Gun
Violence: “Hot spots” patrol in Kansas City,” Justice Quarterly, 12(4), 673-693.
Sherman, Lawrence W., Patrick R. Gartin and Michael E. Buerger (1989). “Hot spots of
predatory crime: Routine activities and the criminology of place,” Criminology, 27(1),
27-56.
Shi, Lan (2009) “The limit of oversight in policing: Evidence from the 2001 Cincinnati
riot,” Journal of Public Economics, 93(1), 99-113.
Skogan, Wesley and Kathleen Frydl (Eds.). (2004). Fairness and effectiveness in
policing: the evidence. National Academies Press.
Smith, Adam (1937). The Wealth of Nations (1776). New York: Modern Library, 740.
Spelman, William. (2000). What recent studies do (and don’t) tell us about imprisonment
and crime. In (Michael H. Tonry, ed.), Crime and Justice: A Review of Research, Vol. 27.
Chicago, IL: University of Chicago Press.
Spelman, William and Dale K. Brown. (1981). Calling the Police: A Replication of the
Citizen Reporting Component of the Kansas City Response Time Analysis. Washington,
DC: Police Executive Research Forum.
Stolzenberg, Lisa and Stewart D’Alessio (1997). “‘Three Strikes and You’re Out’: The
Impact of California’s New Mandatory Sentencing Law on Serious Crime Rates.” Crime
and Delinquency 43: 457-469.
Webster, Cheryl, Anthony Doob, and Franklin Zimring (2006). “ Proposition 8 and
Crime Rates in California: The Case of the Disappearing Deterrent.” Criminology and
Public Policy 5: 417-448.
Weisburd, David. (2005). Hot Spots Policing Experiments and Criminal Justice
Research: Lessons From the Field. The Annals of the American Academy of Political
and Social Science, 599(May): 220-245.
Weisburd, David. and Lorraine Green (1995). “Policing Drug Hot Spots: The Jersey City
Drug Market Analysis Experiment.” Justice Quarterly 12: 711-735.
Weisburd, David, Cody Telep, Joshua Hinkle, and John Eck. (2010). “Is Problem
Oriented Policing Effective in Reducing Crime and Disorder?” Criminology and Public
Policy 9:139-172.
Weisburd, David, Laura A. Wyckoff, Justin Ready, John E. Eck, Joshua C. Hinkle, and
Frank Gajewski (2006). Does crime just move around the corner? A controlled study of
the spatial displacement of and diffusion of control benefits. Criminology, 9: 549–591.
Weisburd, David, L. Maher and Lawrence W. Sherman (1992). "Contrasting Crime
General and Crime Specific Theory: The Case of Hot Spots of Crime." In: F. Adler and
W.S. Laufer (eds.), Advances in Criminological Theory, Vol. 4. New Brunswick, NJ:
Transaction.
Weisburd, David, Nancy Morris and Elizabeth Groff. (2009) Hot Spots of Juvenile
Crime: A Longitudinal Study of Arrest Incidents at Street Segments in Seattle,
Washington. Journal of Quantitative Criminology 25:443-467.
Weisburd, David, Cody Telep, Joshua Hinkle and John Eck. (2010) Is Problem-Oriented
Policing Effective in Reducing Crime and Disorder? Findings from a Campbell
Systematic Review. Criminology and Public Policy 9(1): 139-172.
Williams, J. and Sickles, R. C. (2002). “An Analysis of the Crime as Work Model:
Evidence from the 1958 Philadelphia Birth Cohort Study,” Journal of Human Resources,
479-509.
Williams, K. and R. Hawkins (1986). “Perceptual Research on General Deterrence: A
Critical Review,” Law and Society Review 20(4): 545-572.
Wilson, James Q. and Barbara Boland. (1977). “The Effect of the Police on Crime,” Law
& Society Review 12, 367.
Wilson, James Q. and George L. Kelling (1982). “Broken windows,” Atlantic Monthly,
249(3), 29-38.
Witte, Ann Dryden. 1980. “Estimating the Economic Model of Crime with Individual
Data.” Quarterly Journal of Economics 94, no. 1: 57–84.
Worrall, John L. and Tomislav Kovandzic (2007). “COPS Grants and Crime Revisited,”
Criminology, 45(1), 159-190.
Zimmerman Paul R. (2009) “Statistical Variability and the Deterrent Effect of the Death
Penalty,” American Law and Economics Review 11(2):370–398.
Zimmerman Paul R. (2004). “State Executions, Deterrence, and the Incidence of
Murder,” Journal of Applied Economics 7:163–193.
Zimmerman Paul R. (2006) Estimates of the deterrent effect of alternative execution
methods in the United States: 1978–2000,” American Journal of Economics and
Sociology 65(4):909–941.
Zimring, Franklin., Jeffrey Fagan and David Johnson (2010). “Executions, Deterrence,
and Homicide: a Tale of Two Cities,” Journal of Empirical Legal Studies, 7(1), 1-29.
Zimring, Franklin, Gordon Hawkins, and Sam Kamin (2001). Punishment and
Democracy: Three Strikes and You’re Out in California. New York: Oxford University
Press.